Calculus of IT - Season 3 Episode 3 - Distributed vs. Centralized IT (Redux) - The AI Paradox?

Two years ago, we spent 4 hours and 27 minutes hours dissecting decentralized IT models. Now, AI has changed everything or has it?

In this episode, Mike and Nate return to one of Season 1's most epic deep dives to ask a critical question: Has the AI revolution fundamentally altered how we should think about IT organizational models? (Hint: Maybe?)

The uncomfortable truth? AI simultaneously pushes companies in BOTH directions. It makes decentralization technically feasible (business units can now code, troubleshoot, and build without IT) while making centralization organizationally necessary (governance nightmares, cost explosions, and compliance chaos).

We explore:

• Why the four classic models (centralized, matrixed, decentralized, federated) still matter
• How AI is turning "shadow IT" into "shadow AI", and why it's worse
• The case for (and against) pulling cybersecurity, governance, and employee experience OUT of IT entirely
• Whether every employee should have Claude as their personal IT department
• Why 2028 might require a "verification economy" instead of traditional governance
• The prediction: Most companies will land on a federated model with centralized AI governance and distributed execution

The fundamentals haven't changed, but what HAS changed is our understanding that IT doesn't need to own everything (as if it ever did). The healthiest organizations will recognize what to control, what to influence, and what to let go.

Plus: Why Gemini 3.0 matters (or doesn't), the 5,000-calorie Shake Shack challenge, and whether we're all just one well-crafted prompt away from knowing as much as our bosses.

Episode runtime: 2+ hours of unfiltered IT leadership reality

Transcript

Season 3 - Episode 3 - Final - Audio Only
===

Mike Crispin: [00:00:00] All, all full, totally. Full bars. I love that. 

Nate McBride: Yeah. I like bars. I 

Mike Crispin: love bars. 

Nate McBride: Good. Most, most bars, 

Mike Crispin: absolutely. I'm totally in 100%. 

Nate McBride: I'm trying to remember the last bad bar I went to. I can't think of one. I, I mean, it would have to be like even overpriced drinks I can deal with. It had to be just, I don't know, probably, probably the last, probably the last, it's hard to mess up a bar.

Probably the last hotel bar I went to. 'cause hotel bars are generally not the, uh, not the best. 

Mike Crispin: It's definitely hard to mess up a bar. 

Nate McBride: It is definitely. Um, holy shit, it's week three. Um, it is amazing. I know we've made it, you know, there's uh, 21 working days left in the year. 

Mike Crispin: I did [00:01:00] not know that. 

Nate McBride: Yeah. Let the, uh, sound off sound.

The alarms and the sirens. 'cause that's it. It's pretty much done at this point with that. That few days left. There's no point.

Mike Crispin: 21 days. I can't believe it. You know, we're almost at Thanksgiving and that's, that's enough to be pretty excited about. 

Nate McBride: Yeah. 

Mike Crispin: I love Thanksgiving. 

Nate McBride: Do you like giving you like giving thanks? 

Mike Crispin: I do like giving thanks. I like eating. A lot as well. And drinking and having copious amounts of mashed potatoes. I love mashed potatoes, 

Nate McBride: but mostly giving thanks.

You give, do you give, do you like shovel a bunch of mashed potatoes in your mouth and then give thanks with them in your mouth? I do, 

Mike Crispin: I do. I mumble. I just keep shoveling them in. It's awesome. That's, I love them. I love that. And canned, candied carrots. You gotta get some Candi carrots in there. Some, [00:02:00] you know, some turkeys, some, some great I the stock, I like the low fat.

The low 

Nate McBride: fat, no fat. Candied carrots. Oh yeah, those are, those are, those are no 

Mike Crispin: fun. 

Nate McBride: Yeah, it basically, it's a carrot. Oh it is? Yeah. 

Mike Crispin: I don't, I don't like that. 

Nate McBride: Um, no. Good. So it was a busy week this last week. Yes, it 

Mike Crispin: was. Yes it was. Um, 

Nate McBride: so for tonight, I was thinking we would talk about last week a little bit what happened, surely in the world Absolute of technology.

And then tonight we're talking about, um, we're revisiting, uh, season one, episode 13, decentralized it. Um, and I don't wanna make you go too far back into your way back machine, so I'll see, I'll spare you the, uh, the searching. We, in February of 2024, we had our. [00:03:00] Episode 13 of season one. Okay. Yeah. Yep. Nathan Doyle joined us for that episode, by the way.

Oh, yes. And if you go back and go back and watch it, the three of us were crammed on one side of one side of the table. It was about four degrees outside. And guess how long that episode was? 

Mike Crispin: Four hours and 20 minutes. 

Nate McBride: Did you just make, did you look that up already? 

Mike Crispin: No, I didn't. But I remember that being a very long episode.

Nate McBride: Four hours and 27 minutes. 

Mike Crispin: Wow. 

Nate McBride: So I just remember 

Mike Crispin: getting in my car and driving almost like almost midnight or something. 

Nate McBride: Yeah, that was a long one. So, um, that was, that was one of our banner, banner moments. But, you know, there was a lot, there was, there was a lot of shit to cover, man. There was, um, I mean, poor Nathan, that poor bastard had to freeze for four and a half hours 'cause it was fucking cold as hell.[00:04:00] 

Um, well I 

Mike Crispin: think we, we took a few breaks too, didn't we? Like to, 

Nate McBride: we did for the bathroom and eating. 

Mike Crispin: Yeah, that was great. 

Nate McBride: But, but as I was going through the transcript of that episode, it occurred to me that that was the same episode the week prior to the episode. The tech accord had been announced. Tech accord.

Tech accord. And so I was like, I wonder what's happened to the tech accord. Well, it turns out the tech accord is still alive and, and thriving because not a single member of it has done anything to uphold their promises. 'cause it was, it was all optional. 

Mike Crispin: Are you surprised? I mean, they all got in a room.

They were all excited to be part of something big and then nothing happens. 

Nate McBride: The tech accord. The tech accord. So the te the tech accord is still, for those of you that were season one fanatics. Yeah. Tech Accord is still alive and well. And no one's doing fuck all about helping the planet using AI except to insulate their [00:05:00] homes with a hundred dollars bills.

So that's what happened. You know, what we should do is like, when we have these, we have these throwback episodes, these sort of revisits, we should cover what was happening at the time, February, 2024. That's, that's how far back we have to go for that one. Um, anyway, thanks to Nathan for coming on the episode and dealing with it.

Yes. So we, we focused on, uh, you know, the models, so. Centralized, decentralized, hybrid, federated. We talked about that a lot. Mm-hmm. Yep. So I think we should, I think we should sort of revisit those ideas just to sort of lay, lay the groundwork for tonight. Um, you had made some pretty interesting assertions back then, which I think we should, we could also revisit.

Sure. Um, we, we talked about taking cybersecurity out, making its own function. I still agree that that should be done. Yep. Um, taking program management and governance and [00:06:00] making its own function, and then also taking employee experience and making its own function. But that, that whole talk about employee experience led us into the, um, elimination of IT theory that I have.

And no one likes talking about that. So maybe we don't talk about that. We, maybe we save that one for another episode. 

Mike Crispin: Yes. That's, that's, that's, that's a contentious one for sure. 

Nate McBride: Interestingly enough, I went through the entire transcript and not once did we mention how AI will affect decentralization or centralization of it.

So that will be a good one to talk about tonight too. Okay. I think, right? I mean, yeah. Yeah. The way people think about AI in relation to IT operations for sure. Yep. Yep. Um, 

Mike Crispin: it may actually highlight the need for more it, which is interesting. 

Nate McBride: Yeah. Yeah. So we, we should, we should talk about, I'm make a note on that.

We should talk about that. 

Mike Crispin: I think that's, um. Especially when we talk about like the CPS and the, you know, [00:07:00] the integration and the automation. And it's clear to me kind of talking to a number of people within my company and just, just peers and friends that are in the industry as well. Just that that's just like any other integration type skillset and or data governance, understanding that's something that people who are on the front lines or the on the business lines don't have time to think about or don't wanna think about.

So. 

Nate McBride: Oh, for sure. For sure. Well, I mean, I think also, um, last week we talked about like what the stack would look like for a startup company. I think we should come at that again this time from the model perspective, like what kind of startup model would you build and what, what should you strive for? And with the idea that maybe you're gonna pivot at some point.

But, um, um, talk about that. So yeah, those are some of the things I had, uh, noted for tonight. Um, [00:08:00] and we can kind of cover that stuff. 

Mike Crispin: Sure. 

Nate McBride: I had some other tidbits and I have my special mic question as always. I actually have a couple mic que a couple of special mic questions. Oh, great. Oh great. Yeah, yeah.

Gotta tune up the brain, Mike. 

Mike Crispin: Yes, I'm ready. I'm excited. Good. 

Nate McBride: Um, yeah, so I. I also, I did some research this week on some things and um, just trying to figure out, you know, has anybody has it like from an industry perspective, has anybody changed their position on this? And it's very, very hard to find any literature actually in the last two years talking about it.

Models, other than it models that are reduced because of ai. It's really all seems anyone's able, anyone that it's a, a respectable research nature is talking about. So, I don't know, I didn't really find anything else. I was like hoping to find, you know, [00:09:00] some sort of HBR article or something like, hey, there's been a resurgence in decentralized it, but no.

So it's basically like not much, apparently not much has changed in the model market except to figure out how to use AI to eliminate people. 

Mike Crispin: Do you think some of that is just that more and more people being hired that are in, in business roles, that, that know IT systems, computer systems, they know, have experience in certain systems and it's just kind of an unwritten thing that's happening and it's very gradual change.

Just some functions have technologists, some functions don't. It depends on the company and the people they can hire, that type of thing. Yeah. Might be more quiet and silently sort of moving in one direction versus the other. 

Nate McBride: We should, um, let, let's, when we explore that a little bit too, I just made a note on that.

It's a little bit asymmetric, you know, because we talked about, and we have talked about the editor, creator, [00:10:00] viewer role in the fun in the functions. If everyone is now quote unquote becoming a quasi technologist, um, what does that do for people's rights, abilities, and needs to create new, new content?

Sure. And, and, uh, and we're gonna cover this in another episode, but we can probably chat on that tonight a little bit too. Sounds good. And then, um, but I really think, you know, like getting to the, I think the meat of the competition will be about, um, about the AI parts, uh, what's, again, what's changed, if anything.

Um, and then this actually comes back to last week's discussion because, well, if, think about it, if my IT department's gonna be decentralized and have human beings and AI agents, um, then we come back to the identity question again. 

Mike Crispin: Sure. 

Nate McBride: Um, so [00:11:00] anyway, that's about all I had. That's probably gonna be a little bit of a longer episode, but, 

Mike Crispin: okay.

Okay. 

Nate McBride: Sounds good. Okay. Well got your sleeping bag and your, your booties and food. I got everything. 

Mike Crispin: Get everything ready to go. I got some. RINs and my candle's burning and my heater's on. Okay. Ready to rock. 

Nate McBride: All right. My heater's on too, though. It's not really working as efficiently as I would hope at the moment, so I'm still, I'm still freezing, but, um, okay.

Well, I thought we would actually do a new feature tonight. Okay. Which is to hop into the, um, calculus of it, mailbag. 

Mike Crispin: Oh, okay. That sounds good. Let's do it. 

Nate McBride: Yep. And just see what, see what's been coming in the mailbag,[00:12:00] 

Trance Bot: the calculus of it,

season three,

verifying this identity.

Sometimes you 

just have to take it.

Sometimes you just have to take it

because it's season three divided autonomy

enemy.

The calculus of it.[00:13:00] 

Nate McBride: So, um, I'm gonna reach deep into the bag here, just reaching deep. What kinds of stuff down here? Who 

Mike Crispin: knows what you'll pick out? I mean, 

Nate McBride: in the, in the s You gone through thousands 

Mike Crispin: of messages here. 

Nate McBride: See, Viagra, Viagra, Viagra, pee leaks. Erase pain. Um, oh, 

Mike Crispin: erase pain. That's a good one. I like that. Yeah. Sticky breath.

Nate McBride: You get that too. Stinky breath. Col. Cholesterol, blood pressure. Uh, let's see. Penis enlargement. What's a pice? I have no idea. Theis. That one's, that one. That one's, that one's for you. Sorry. Um, ah, let's see. I don't know. Know what that 

Mike Crispin: is? 

Nate McBride: Hoof evidence about hoof. Phil Collins murder, blah, blah, blah. Uh, I don't know what that one, [00:14:00] they want some Bitcoins.

Uh, let's see. Oh, we got one. Um, wow. Zs so many emails. Those darn cookies. Zs, where the fuck did Z Zs sounds? Z. Where did Zs come from? I dunno why that just came in my brain. Where did Zs come from? Hold on. Zs, I feel like that's like a cartoon word. 

Mike Crispin: Is it? Isn't that like a place you can buy used instruments or something?

Nate McBride: Uh, well used, used flutes. Well, no, it was used in Scooby Doo. 

Mike Crispin: Oh, alright. I thought you were talking about the the store. Yes. Sounds 

Nate McBride: was also an English anarcho punk, post punk band. Punk post-punk band. And then it actually means an archaic exclamation of surprise or anger, a euphemism for God's wounds. You can also refer to an English anarchist [00:15:00] band.

Right. So I don't know why I just said sounds or it came to my brain. I was having a Scooby-Doo flashback moment, but okay. Back to the male. Back. Jesus Christ. Um, back to the mailbag. I want some money. I've lost some money. You are picked to no one else. Okay, here we go. A question from our loyal, a loyal listener.

Our, our loyal listener. 

Mike Crispin: Our loyal listener. 

Nate McBride: Listener has asked a question, okay, what tools do I like from markdown in Gen ai? 

Mike Crispin: Um, what tools do I like? I'd say obsidian. I know I keep saying that word, but I use that for, and, and I like giraffes as well, which is an apple, apple only product. But I use obsidian for a lot of the markup work that I do.

So the other 

Nate McBride: one that, there's another one that's Apple only. It's uh, deck. Deck set. Deck [00:16:00] set is Apple only. Yeah. Uh, and, and that one's, I like that one. I've used Marp. Mm-hmm. mar.app. Um, that's a good one. Um, in fact, that's my, that's kind of my favorite. I have it open right now on a tab in and chrome, um, sly dev, but that's more like a developer focused markdown.

Sly, yeah. Dev. Pretty unique use of TDS there. And then if you're doing like, I, it's more like, it, it's not so much markdowns as markup as reveal js. So reveal, 

Mike Crispin: are you talking about markdown text files? Are we talking about like mark marking up documents? 

Nate McBride: Taking, taking the markdown that comes out of an output and then Yeah, like a repre repurposing that.

So like you ask, uh, an m an 

Mike Crispin: MD file, right? An MD file. 

Nate McBride: Yeah, yeah, yeah. You ask Claude for a presentation, it gives you markdown and then you can bring that into Mark. Yeah. I, I would to actually make a PowerPoint. [00:17:00] 

Mike Crispin: I am, uh. I'm all in an obsidian because I mean, that does you have that, then you have all the add-ins, you have all the community add-ins and everything else, if you want to use that.

And, um, can they, I really like that we can canvas those too. Like you can take all the MDs and you kind of put them up on a wall and see 'em, like sticky notes almost. I, I dig that. 

Nate McBride: Mike. Mike, do you, do you, do you like obsidian? 

Mike Crispin: Love it. I can't say enough good things about it. Tell Yeah, I wasn't 

Nate McBride: sure. I wasn't clear to me if you, if you liked it or not.

I couldn't tell. Yeah. 

Mike Crispin: It's one of the few tools that like, I haven't like flipped off to something else pretty quickly. I've been on it for a couple years now and I'm like, this is fantastic. And you know how fickle I am. I jump around from 

Nate McBride: you're, oh, you're quite fickle. I'm f 

Mike Crispin: fickler, I'm fickler, I'm a fickler.

You're 

Nate McBride: a fickler. You're a fickle stein. You can say that once more. Fickle fickle stein. Fickle fickle, 

uh, fickler by the way, based on all the [00:18:00] spam loom. I have to wonder if like someone out, somewhere out there, there's still like an exchange 2003 server that someone forgot about. That's literally every single spam bot in the world on it.

Mike Crispin: I bet there are thousands of those. Who the fuck, fuck is cooking these links? 

Nate McBride: Really? 

Mike Crispin: They're probably running on a, on an old VMware server somewhere. 

Nate McBride: I mean, there's, these aren't even good. They're, they're so bad. Like they're, I, I saw these emails already in 2004. I, it's coming back

like, oh my goodness. I want to quote and would like to know your availability so I can send you the necessary documents in order to exchange a transaction. I don't even know what that means. Five more signs you need to send. Neither do I. Oh, here's good news. Here's my purchase order for the things I didn't order.

Thank God. I swear to God, there's, [00:19:00] there's just like rogue to, you know what would be funny would be to, to like get 2003 exchange servers and just open up his honeypots and just let 'em go, let people go to town on it. 

Mike Crispin: Absolutely. 

Nate McBride: Just put it on my, my, my company. My company, my building's wifi and let it sit there in a corner all day.

Mike Crispin: Just, I, I think it'd be great to just plug, plug a, just like a spanning tree loop and plug that thing in there and get it going. 

Nate McBride: Like, like, like start, start it clean with nothing on it and 

like, we'll just let it run for a couple days and see how infected it gets. That'd be pretty funny. You 

Mike Crispin: have all the good ideas.

I know, I do, don't I? 

Nate McBride: Well, I had another idea. What's that? So I I, I'm over kind of over most of my running injuries, so I started running again and yeah. Um, thank God for everybody else. Honestly. Everyone's taking a big breather off of that period of time. Um, I was thinking like the gen AI [00:20:00] industry is like the, like the eighties music industry.

Okay. So, so we, we play Sirius XM eighties on eight all the time in my office. And, um, every now and then you're like, oh my god, this a cool song. I like that song and this song and Banana Ram and all these other fucking bands, right? Every now then there's this song where I'm like, I've never heard this song in my life.

And it was like number 17 in 1986. So I'm like, I never heard this song. This is a bullshit. This is totally made up. And then they're like, that song was a collaboration by, uh, you know, Don Henley and Eric Clapton and some jackass. And it got me thinking like, gen AI is pretty much a direct corollary to the eighties.

'cause, 'cause yeah, yeah. It does some cool things, but for the most part it's just people are just creating shit everywhere. These shit companies that are like, not even one hit wonders that everyone's gonna be [00:21:00] around. Like, did you ever use that product? I was like, I've never even heard of it. There's so many of them.

Tell me how you really 

feel. 

That's my analogy. Oh my god. Hold on. There's a big giant fly on my monitor. I 

have to pull. Oh my gosh. 

I don't even know why he's in, in here. Get outta here. It's winter, dude. Um, yeah, so that was my observation. I was running and I'm like, you know what? Fucking ery AI is the eighties, just a bunch of garbage.

And you just like want to get to the good song know. So, you 

Mike Crispin: know, you don't think that was anything good in the eighties that's worth keeping or just it was just, you were just saying there was an overflow of just tons of songs, tons of crap. 

Nate McBride: I'm, I'm saying like now I'm enough of a connoisseur of eighties, one hit Wonders and eighties music.

Yeah. And I have been from, I mean, my decade right, from my, from most of my life that I can hear things now and immediately you like know the artist and whatever. But the point is that every now and then [00:22:00] SiriusXM, for whatever reason, will play a song that no one ever liked, ever because it just existed in the eighties.

And then it turns out that that person had like a, a contract and that record sold a bunch of copies and they had a tour and you're like, Nope, nope, nope. That's impossible. This is such bullshit garbage song. How is that even possible? And I can cite many of these. But, um, yeah. So anyway, that was my, that was my, my deep thought during my run today 'cause um, that a 16 z uh, thing, that they're basically giving a million million dollars to anybody who can.

You don't have to create something, you just have to talk about creating something and they'll give you a million dollars and in case it works, um, I was like, oh my God. The amount of shitty apps that are gonna come out, or they already gonna do is just 

Mike Crispin: tell it. Right? Just tell it. 

Nate McBride: Yeah. I'm thinking of an app that [00:23:00] when you ask it a question, it doesn't answer.

Mike Crispin: Ooh. 

Nate McBride: It responds with dot, dot, dot. And it leaves like a, you know, like when you, when someone's gonna text you and they don't finish a text and it stays as the dot, dot dot on on, and you don't, you don't know what they're gonna say. Yeah. I'm gonna call it the suspense spot. And it just leaves you, it leaves you in permanent suspense, and then one day you don't even, and you just keep 

Mike Crispin: typing and typing in and hoping that something good comes 

Nate McBride: back, keep rage typing, and then, then one day it actually responds spontaneously to something you've said in the past.

You don't even know what 

Mike Crispin: I, hold 

Nate McBride: on. 

Mike Crispin: Oh my God. 

I know that flies are crazy, man. You can't get them. I had one in my office this morning and I could, it was, I could not get guy, this 

Nate McBride: guy was drunk. He was so big and fat and slow. I was, that was an easy grab. All right. So anyway, that was the thought I had. So, listen, I know you're itching to talk about, well, I know you're itching, but also itching to talk about what [00:24:00] happened yesterday.

New the Gemini three. What's a news? Oh, that, that news, that little chestnut. Well, it's so funny because just last week, what were we talking about? 1 billion models that have come out 

Mike Crispin: in the last 

Nate McBride: year. 

Mike Crispin: Yeah. And who cares about models or why should we care about models? Right. So, 

Nate McBride: so before you go into your Gemini 3.0 spiel, let me ask you a question.

Why did they bother, why did they bother to release 2.5 if they knew 3.0 was coming out? And I, it's only partly rhetorical because in truth, yeah. The only reason they released 2.5 was just to have something, in my opinion. 

Mike Crispin: I, I I, I agree. It's, that's what it always is, is kind of the, we gotta keep ourselves on the map, we gotta keep ourselves relevant.

And if they feel like there's enough of a push forward that they, they're gonna put it out there. And if they can put marketing around it, they will. I mean, look at the recent, um, GPT 5.1 and Grok 4.1 and it's like, can you put a [00:25:00] little package around these things and create, uh, some, get some attention, get some, you know, some stir up to keep people focused on your model.

Yeah, there are some good things about all those updates, but are they hugely newsworthy? Probably not, but some of the things. That I think could be even, or should have been even more newsworthy was when Claude came up with MCP or when you had the Gemini user interface, you know, got built into Google Docs, you know, stuff like that.

There's truly use good user interfaces, um, are really important. But with Gemini three, it's kind of this full multimodal, um, approach that's in open AI and others as well. It's, I think it's why there's so much news around it is 'cause so many different parts of the wheel were improved and it's harder to compete, I think against when you have those multiple modes.

But the other thing is I do think be, maybe [00:26:00] it's because there's so much press around it or just there's a lot of noise, is that I think there was largely this belief, at least in the last few months, just to saw the bubble popping and everything. Yeah. That there couldn't be a significant improvement.

Everything was gonna be, at least for the foreseeable few months, it would be these little point releases that were gonna happen. And I, I think the, the Gemini three release either, well, they just have great marketing or it's, or their, their benchmarks are really that good. Um, this gives I think investors more, more, uh, push to, Hey, we haven't hit the ceiling yet.

It just broke through another barrier and let's just keep putting more money in. So I think that's. Sort of what's happening with it. I have been using it and I think it's, it's pretty incredible. But you talk about where like OpenAI and, uh, Claude are and how well they can integrate with your workflows, that's still something that Google has to work on.

So [00:27:00] in terms of this extensibility is still, um, still requires a little bit of work. And, and you and I both know that a month or two from now, we'll go around the circle to the next one. It's gonna be the greatest thing in the world and the next one. So to your point in the last episode, why do we, why do we care if there's a continuous cycle of these things?

And I, I, I think some of it is if I think you're a technologist and you're, you're hopeful and you are optimistic about the future, you're always gonna wanna know. And there's these big leaps. But as a, probably as an IT leader, as a CIO, you need to pick the right platform that is gonna fit in with your org that follows the right principles and policies that you have.

'cause it's, all of these are gonna catch up unless one of them gets acquired and shut down or whatever. So it's hard not to pick one of the big three or four, uh, or go totally open source and really risk risk things. But I think that right [00:28:00] now it doesn't, it doesn't matter what you choose. It's, it's gonna be, they're gonna continue to evolve.

But I do think it's interesting to know if there are some major. Frontier model breaks that you sort of know about them, even if it's just a precursor. Your how your, what you've chosen to invest in and build in your company will soon be affected by whatever the next wave is. What you can expect that happened in Gemini three.

Right now, you're probably gonna see Chacha PD six if you're on that platform, or Anthropic probably more likely to grow their integration and ecosystem. Their, they just had the deal with Microsoft, they announced. So I think there's two fundamental different approaches between Anthropic versus open AI and um, and Gemini where, um, open ai, you know, was largely connected to Microsoft.

That seems to be breaking apart a little bit [00:29:00] and OpenAI is a little bit, yeah. To, well, totally, I guess, right. Starting to move off and b, really its own org and it's under a ton of scrutiny right now, I think. And a lot is probably rightfully so. And then Google just had this big hit, I think, I mean, this is, everybody is talking about it.

There's a lot of excitement, I think, in the AI community and in the technology community about it. And maybe it brings more people into their ecosystem. So, who knows? 

Nate McBride: Well, we don't, we don't get to the walled garden episode until episode 15, but I tried to time that episode, which will be like, sort of late March.

I tried to time that episode so that we would be fully, so that every single, every single, uh, one of the bigs would have their browser officially sussed out. Yeah. And it would be, um, you, you'd be working inside that particular vendor's browser to do your work. So, you know, we already know OpenAI and Atlas [00:30:00] are moving towards the walled garden approach.

Sam Altman has said as much, but, uh, perplexity has, comet Opera has neon. Microsoft is doubling down on Edge AI and Google has Chrome. So we, we can, we can, we can, if we start to like project a little bit forward. Yeah. I think we can see where we're headed. The, i, the, the, the idea of like, it, it, instead of being a 2.6 release into 3.0, it's pure marketing, you know, sort of thinking, um, yeah.

Because Google's very good at using single point releases and not, uh, not full version releases with most of their products. Yeah. So, so there's that. 

Mike Crispin: I think it's interesting to see that ecosystem versus the API model, right? Yeah. And that's, that's, that's gonna be the, I think there is sort of the two models right now is you're either.

Uh, gr uh, X AI and GR API or anthropic API model or your [00:31:00] open AI ecosystem, Microsoft ecosystem, Google ecosystem, because it's clear that open AI is gonna have an office suite by the end of the year, early next year. It's clear that they're gonna be building the same things. They see the writing on the wall.

So they are going to try to compete and be just like Microsoft and be just like Google. Um, and then even it 

Nate McBride: reminds me, it reminds me of like the Facebook's attempt to get into the, I remember that market and all that shit. I mean, I think it's a failure. Like, like if, if Google can't get through the 98% market share of office, how in the shit is OpenAI gonna do it?

Um, it's that it's gonna be the walled garden. And, and of course I think that would be very interesting to see if Google makes a final and official bet someday to say, oh, you wanna use all of our amazing products, you have to come into, uh, Google Workspace. I mean, how, how badass would that be? But they have to get everybody so hooked, I think, and be so [00:32:00] far out in front that that's gonna be the key.

I don't know if that's possible or not. I, I do think 

Mike Crispin: there's a risk of, if you're really worried about lock and locking inside of Google is not necessarily the best thing. 'cause now you're, well, you are totally, you're I and I love Google, but it's just that, that, that's a pretty sticky platform that though they have a great exit strategy, you know, takeout and all those tools.

It's kinda like, all right, now you're really, it's really gonna be larger, how you lose some autonomy. Yeah. How do you 

Nate McBride: get a license for, for alaw? You, well, you buy it from Anthropic. How do you get a license for, um, chat, GBT? Well, you buy it from open ai. Like you, you buy the license, you have the access. How do you get a license for Gemini?

You have to buy Google Workspace. That's right. And that's, can you, can you host 

Mike Crispin: Gemini on, on Amazon or on a third party? That's a way big 

Nate McBride: fucking different deal. And that's, uh, that's right. Unless you use perplexity or PO for instance, and choose your engine, ultimately that's the only one you can't use.

And nobody, nobody's gonna use a consumer grade gen AI platform at a [00:33:00] enterprise level. So 

Mike Crispin: Perplexity is an absolutely fit. I won't go, I won't say too much. I know you, I can't set about perplexity, but it's not an AI platform. It's a tool. It's a search engine. An aggregator. Yeah. Yeah. It's a search engine.

So like if you, you it's, and it's great. It works really well, but it's kind of like going beyond, um, a, a search engine and trying to extrapolate other models. Yeah. Like you, like Poe, right? I, I agree. But when you start talking about coding and agentic work and, um, cybersecurity and everything else, I think that they, they're a good on-ramp.

I think they are a fantastic starting point for a company to look at what PO or perplexity before you get into needing all the guardrails that you'll need with, uh, one of the real platforms. 

Nate McBride: Don't look at PO though, 'cause PO doesn't have enterprise licensing, but yes. You know what's the, let me ask you a quick, here's a pop quiz.

What's the number one thing people will cite as their reason for using [00:34:00] perplexity, uh, references. Exactly. Which you can simply use the cite all sources line in any other gen AI prompt and get the same output. Um, or use deep research. So it's like, uh, oh yeah, that's exactly right. That's the answer. Every single person gives you perplexity only 'cause they're unwilling to type one single, uh, seven word prompt.

So, um, 

Mike Crispin: the only other thing I'll say is just speed. I mean, it's by far faster than all the others. It is not even, it's not even close and accuracy in most up-to-date information. It's never confused as to what to use to pull up. Sure. Like you can search on Gemini and or Chacha pt and it's still things that sonnets three, five, you know.

So it's like those are the things that, you know, even if you tell us to go to the web, the most UpToDate information's probably gonna come up out of Plexy most of the time. Whereas I always have to, I have to reprompt. Constantly reprompt, uh, any of the other big models and say, look, [00:35:00] I want up to today's date.

I'd like you to look up today's stuff. You know, there's that force 

Nate McBride: multiplier in effect for you right there to type it on the 

Mike Crispin: sentence. Nate, I don't wanna type another sentence. Come on man. Man, you, me. You 

Nate McBride: mentioned 

Mike Crispin: the bubble 

Nate McBride: earlier. So Microsoft is basically like, okay, well fuck the bubble. We're gonna go ahead and create an agent that manages your agents.

You read that news? I saw 

Mike Crispin: that. I saw that too. 

Nate McBride: Yes. Like, okay, I love it. In case you haven't doubled down already on us and, and hope. Mm-hmm. Now we're gonna make you triple down on us. Oh my God. Who's gonna buy into this shit? Um, it's like a dream 

Mike Crispin: within a dream without a, within a dream. We're in inception here.

Oh my 

Nate McBride: God. Oh, please, dream stacking. Don't do this. Don't buy it. Um, and nobody, nobody at Microsoft really knows how to articulate it. If you read all the nadella sound bites and all the little things, it's just all, no one knows how to like, use actual nouns and verbs to describe it. It's pretty, pretty interesting stuff.

But, um, I had another question for you. [00:36:00] Switching gears again. Yeah. Uh, I asked this in the Slack board and I think everyone's too embarrassed to answer it, so I'm gonna ask you, you gonna ask me, are you, so are you doing a pen test in 2026? A logical pen test. 

Mike Crispin: I, yeah, I'll, I have, I already have done some, but I'll, I can just say that we're doing one.

I probably won't go much more deeper into that. Yeah, yeah. But yes, I mean, yes, we do them. Yes. So, 

Nate McBride: yeah. Yeah. So we do them too. And I was thinking about the 2026 pen test model. Like what, how I'll want to sort of framework it Oh, 

Mike Crispin: yeah. With everything that had with philanthropic, right? They little story. Yeah.

Well, not 

Nate McBride: only that, but with my, in my 2025 pen test model, the company I used, AI was never mentioned. There was no, and that was 2025. So that was mm-hmm. Um, in Q1 and 2025, I'm going to 26 now, and I'm like thinking to myself, well, do I just continue to do a logical pen test, um, structure the same way I [00:37:00] used to and ignore gen ai?

Or do I now have to include that in scope? And here's the thing is it wasn't even offered to me by, and this is a very reputable firm. It wasn't even offered to me in Q1 of 25. So I'm wondering very much how it's gonna go down in 26. You know, will my vendors offer it? Will I need to ask for it? And how will I find a vendor that understands how to do it 

Mike Crispin: Well, I think there's the, the two sides of the coin are that.

Good. A good pen tester will be, is using it already, and you don't even need to know that they're using it. Like if they're, if they're writing or scripting something in open AI or chat GBT to infiltrate your network, there's not much you can do about that. Other than that they're using, they have more resources to create an attack against you than before.

But I do get concerned about, it's like when you do like the internal pen tests, which I think are even more interesting because this, this, [00:38:00] and this is one of the things I think I've been thinking about a lot is, okay, let's start talking about the internal pen test. Someone is, uh, upset at your company.

You have a disgruntled employee. You have someone who's either or infiltrated the network. What tools are you giving people to use, to use to do ai? You know, if you're, if you're someone who is giving people a, you know, $150, $200 a month license to run cloud code in your environment, then there's a really good chance and they have admin rights in their machine, then they, chances are they're gonna be able to write an application to circumvent your security from the inside, right?

No matter how much you do, that to me is a little more dangerous than the outside because you, there's, there's some things that are gonna be outta your control, and if someone's gonna penetrate you from a, uh, external perspective using these tools, they're probably doing it already or trying to do it. So, if, if they're using Coli Linux and just going through and running [00:39:00] the traditional commands, that, that's a good checkbox.

But Chizar, um, a lot of these MMSM SSPs, they could do 25 pen tests instead of one at a time if they write this code. So there's gonna be a, it's, it's gonna be in it for them to, uh, be in the money for them to learn to use some of these automated attacks. I, 

Nate McBride: I agree with all that. My question, I guess was more around, um, how do you know the vendors?

Are you, are you going, are you going to have them test. Are you going to include that in the scope of testing for yourself in 2026? 

Mike Crispin: Yeah. Oh, I'm, I'm, I'm, I'm gonna ask them like that they use the most modern possible techniques. Absolutely. Okay. Yeah. Yeah. 

Nate McBride: A hundred percent. Yeah. I'm thinking the same thing.

I just dunno how to, I, I'm still thinking today, I'll probably sort this out later, but still thinking about how I'm gonna narrow it down that scope. Like what specifically am I gonna want them to test, uh, from that perspective. Um, just something I was thinking about, 

Mike Crispin: there's still sort of on the outside looking in, unless you're gonna give [00:40:00] them internal access to do certain things.

And that's when, that's when I think you can get more particular, because it should be the same pen test externally as you have now. It's just that you would expect them to have better tooling in 2026 based on everything that's happening than they do now. And I think as part of your engagement with them, you can ask and say, what are you using?

And they should be telling you exactly what they're using. 

Nate McBride: What are you using? 

Mike Crispin: Damn it. Who do you work for? What's the problem? But, um, they'll be pretty, they'll pretty straightforward. There's no secrets probably there. And you're, you're under NDA, they'll probably show you some of that stuff. 

Nate McBride: Sea tech astronomy, man.

That's right. Too many secrets. Too many secrets. 

Mike Crispin: Greatest movie. That's, that's how I get into it. By the way, did I ever tell you that? Yeah. That move, that movie is what just got me completely into computers. Like I saw that movie in the early nineties and I was hooked from then on. I. Started [00:41:00] building computers, started every, that movie did it for me.

Nate McBride: That's been the name of my cybersecurity committee since 2008. I always, every company I love is cec, CTEC, astronomy. I love it. What is that? I'm like, here's the link. Go ahead and watch it. Um, you, you mean not cooties, 

Mike Crispin: cooties rat.

Remember on the anagrams that's in the movie.

Nate McBride: Um, all right, so tonight we're talking, we're revisiting a topic. We, we talked about in 2004, not 14. Jesus Christ. Um, we haven't been podcasting that long, although it would've been cool to start in 2014. It would've kind of been like at the front, the front of the, of the vanguard. You know what I mean? Yes.

It would've been like, why are these two knuckleheads talking about ai? That's not even a thing in 2014, but we would've known anyway in [00:42:00] 2024, February, to be exact, we released episode 13 of the first season of the Calculus of IT podcast. We were young then. We didn't really know what we were doing. We were, we were, um, we sat around the table very young.

We drank a lot of, a lot of things, and generally we had pretty okay audio, but it was hit or miss for the first few weeks. Yeah. And on that particular night, we talked about, um, distributed versus centralized. It. And, uh, it was a topic that I had really wanted to dive into and we had special guests and we did all this great stuff anyway.

Um, 'cause now we're here two years later and in a lot of ways the landscape has not shifted. 

Mike Crispin: Not at all really, 

Nate McBride: other than maybe companies have shrunk and downsized because of industry, uh, problems. Certainly in life sciences. People had to rethink their structures, but also because of this stupid shitty thing called gen ai [00:43:00] that's affected all of us.

So we're gonna kind of jump back into all that. Uh, as I said at the, at the pre-read, I wanted to just quickly go over the four major models we discussed back then, um, and jump in and tell me if I got all these right. We talked about decent, uh, fully centralized it, which was a single IT group and everyone reported to into IT leadership.

So either the person at the top or the person below them. All decisions came from the top down. All budgets were controlled top down. And then, um, it would, it worked well and sort of like places where you needed consistency and it did not work well when the business was moving faster than it that's fully centralized.

It very sort of like rigid police state it basically. Sure. Then there's a matrix matrixed it, which is kinda like the hybrid model. Which is, uh, is a central IT ID idea [00:44:00] like that there's it and it's a thing and it's centralized and there's, there's a, there's a thing, but there was distributed execution.

There was people that were in certain functions, maybe they weren't necessarily an it, but they had it like, uh, responsibilities. But there were actually it bullets. Some, sometimes it could be in a different functional line, but somehow you had the two working together and it worked well. Um, mostly like a combination of dotted line reporting and solid line reporting.

Mike Crispin: Mm-hmm. 

Nate McBride: Um, we, we do, we do this pretty well today at alio. Uh, it works kind of medium sized companies and like, especially like geographically distributed companies. I talk about this in the, uh, the autonomy paradox book, and then it, it fails well whenever there's, um, confusion over who has the final say.

So that's matrixed it, it's hybrid and there's fully decentralized it, which is that, uh, there's an IT leader in every business unit. So obviously you would have to have a [00:45:00] company that had defined business units. Um, and then each unit would have its own IT leader. And the central, uh, IT leader is kinda like a visionary only.

They're suit, they have absolutely no tactical operational objectives. They're generally, um, golfing and going to boondoggles all the time. Uh, showing up at garden events and, um. They don't really do much except have an idea and they get everybody else to drink the Kool-Aid. Everyone else is doing the work.

Um, it works in large, very large companies that are generally mature. And, um, it fails whenever you have inconsistency or inability to leverage economies of scale, especially when you have, um, multiple platforms to do the same thing across business units. And then the last one was federated, which was simply shared services.

So you had, um, like cybersecurity, governance, [00:46:00] operational it, sorry, cybersecurity and governance or centralized and operational. It is distributed kind of like a hub and spoke model. We talked about that in 2024. Um, it works with companies that have very, very strong operational discipline. So COOs that are sort of like super tight, they have their shit together, everyone's marching to the same drumbeat, and then they fail wonderfully whenever the hub becomes a bottleneck.

And everyone's like, I don't know, Mike does, Mike does that. Mike does that. Mike's like, no, no. Go talk to Mike. Hey, me, it's Nate. Go. Go talk to Nate. I just talked to Nate. You said talk to Mike. So those are the four models, right? Remember, remember talking about that? Yep. It's not my fault, right? Yeah. So when you're, when you're not fully centralized, it's awesome to be able to be like, I don't know.

You gotta go talk to Mike when it's fully centralized. You, you can't talk to Mike 'cause it's Mike. Mike is at the top of the chain. He's [00:47:00] the only one. Um, so, all right. Did you, you want, did I get all those? Did I nail that? I think so. 

Mike Crispin: Yep. 

Nate McBride: Oh, nailed it. I love it. So I was looking into our, I was looking through our transcript, man.

We, we said some funny things. Um, maybe we should just re-lease that whole episode or some of the highlights of it. It does. I wonder, I wonder if, uh, if I put it through the notebook, LM box, what they say running through Gemini. Well, they, they, they, they do about that episode. Yeah. There was a, there was a lot of uses of the word fuck in that episode.

Um, yeah, there was 

Mike Crispin: a lot of swearing. Yeah, I remember that. I was swearing. Luckily, I think the audio was a little bit off, so you couldn't really, couldn't hear most of them. 

Nate McBride: So, uh, yeah. Yeah. Actually the audio 

was pretty good in that episode. Was it? Yeah. I, I listened. I was like jumping around and I was cracking up.

Um, man, we were young two years ago, [00:48:00] so Sure thing. Uh, sure thing. So you said at one point, even in distributed centralized models or dis distributed decentralized models, cybersecurity must be centralized. Yep. You need guardrails and rules. The same is true for governance. There has to be a hub function.

Do you still agree with that statement? 

Mike Crispin: Yeah, I would say so. Okay. Yeah, I think there to be some, somewhere where standards live, whether those are cybersecurity standards or they're some sort of technology principle standards. And yeah, maybe more so than ever being that we're in this AI world where yeah, still garbage in, garbage out.

There's gotta be some function that's focused on making sure that it's good stuff in good stuff out, and they're focused on that predominantly. 

Nate McBride: We, um, we had talked about, I remember Nathan had [00:49:00] made a point about this too. If you take cybersecurity out and if you take, um, project management slash governance, you know, just the general idea of like adulthood out of it, and then you take out employee experience and make its own function of what's left and it was, you have operational it, like mm-hmm.

The nuts and bolts keep the lights on it is all that's left. Yeah. And um, so, but each one of those groups would either need to have its own leader in decentralized it. So like if you have this play experience group, they have to report to somebody. You have a, a cybersecurity group, they have to report to somebody who reports to somebody and like so on, so forth.

Right. So then you, who's the I, who's the IT operations group report to. Yeah, because that's now considered central there. Sorry, not centralized it and decentralized model, but that's the, that is it by, in the book version of it. 

Mike Crispin: Yep. 

Nate McBride: We never, we never [00:50:00] answered that. Yeah, because I was looking through, and I, we kind of left that as an open question.

Mike Crispin: Yeah. It's all, and it's all over the map. Right? In terms of the reality we're in today, where does it really report? Does it report into, uh, a risk-based function like, like legal and sometimes your COO is very heavily based in the risk space, whether with the quality and technical operations teams. Does that make sense?

You 

Nate McBride: still under the belief that CISOs should not be reporting to CIOs? I, I mean, I, I still believe 

Mike Crispin: I, I am, I am. I I still think that's a very good thing 

Nate McBride: that they do report to CIOs or that they No, 

Mike Crispin: that they, they do not. They should be a peer. I think, um, it should report to the CCO, the CEO, similar, I would think of it almost in the same breadth of the chief legal officer, like if, right.

It's almost that parallel. It is very interesting that. Just two people have focus on risk just, just [00:51:00] because it's digital and it related that it doesn't get that seat Unless you're a product, you're a software company. It's have a real digital you a real ct 

Nate McBride: o Yeah. It's a digital at bottom, bottom layer though.

Up top. Yeah. It's risk and governance. I mean, it's, it is, 

Mike Crispin: it is. But that's what I mean, like, I think still very much looked at as a technology role. Yeah. And I think companies that are software product centric and life sciences companies that are, that have a, a, like a real ct, A CTO, not a chief technical operations officer.

Like a, like a, um, tech operations person. Yeah, tech, tech 

ops lead. 

But a real technology person, they usually have a separate CI o and maybe they do have a CIO, but they have a CTO and a CI o and they may have a CIO or some, and a lot of times the CTO will have a chief of staff that helps to manage the infrastructure and operations part of it.

So it's, that's, I think, I see that's the co the company is a technology [00:52:00] focused company and I think still in our, at least in the smaller pharma space, it's still such an upstart and an upswing that they're really just looking for operational efficiency and we stay in the same model, probably more likely.

The hybrid model that, uh, model number two that you mentioned and that we talked about the first season. To me is the most common because of the blend. Because it really depends on, and I think our, again, we're talking small pharma, a lot of people are hired based on who they know. So you might have someone who knows a very strong technical person who's in quality, but the, the VP of quality at another company doesn't have someone like that at their last company.

So they rely on it to help them. And I think it really depends on the people and the network that leadership has these small companies. And that's why we see such a distribution of, uh, hybrid roles and, and a hybrid model. Someone has an expert in regulatory [00:53:00] applications that the head of regulatory might just bring them in and they partner up with it and they go, they don't have that person.

It tries to help out as best they can, or they really fight for that role, or why they would, I'm not sure, but they, they would fight for that role. Some will, and I think depends on the CIO and what level of control they wanna have, uh, over and maybe their expertise in certain business functions that they wanna push.

So I think it's a really, in a small company, it's like a, it depends. Sure. Because we're, we're hiring people we know, um, in a lot of instances, small industry, 

Nate McBride: it is small. I know in my entire career of knowing CIOs and who they reported to, I know of only two instances ever. Where A CIO reported to A CEO.

Mm-hmm. They com they most commonly report into A CFO. Um, and in a few cases they report to a COO if one exists. And COOs are rare and rare these days it [00:54:00] seems. Mm-hmm. But they don't report into CTOs for sure. They don't report into c CDOs and CTOs in life sciences definitely have more of a, um, they, they, when you hear CTO in life sciences, your first reaction is that person runs technical operations.

Yeah. It's manufacturing. Yeah. When you hear CTO in any other industry, it is the person who's running the development side of it. They're generally running the development shop. So then the 

Mike Crispin: product, then the product manager. 

Nate McBride: Yeah, the product manager. Exactly. We don't the product lead, we don't really have that in life sciences except for those companies that have gone digital route for its internal development.

Um, so, so has anything changed in decentralized it? Like I'm, I'm, I've been thinking about it and, you know, I was taking some notes about this. I mean, I don't think really anything's changed. I would still, I would still run it the same way. 

Mike Crispin: Mm-hmm. I don't, I don't think anything has changed. And I think it's more likely that it's a pe, it's a piecemeal approach that, [00:55:00] you know, it really depends on the talent, um, that's brought in, in the leadership of each function and, and who they want to hire.

Unless you know, your, uh, head of people or head of HR is. Is really setting a certain operating model. Sometimes they come in with a, an operating model in which certain skill sets are gonna be arranged. And yeah. Then you may have a more of a voice of how to distribute it. But I think if it's kinda leave it up to the functions, then quote unquote that, um, it can very well be a hybrid model.

And I, I can't say that it, you know, in a lot of instances that it hasn't worked. I mean, it seems to work pretty well at a small company, as long as you can maintain the relationships and that you've got a good, you're part of that interview team when those type of roles are hired, um, I think you put yourself in a good position.

Nate McBride: Yeah. You're gonna, you're gonna get what you're gonna get. You think you need to, I mean, for decentralized it, I [00:56:00] think the same holds true that we talked about. You would need to have very, very strong security. Yep. So if you're gonna de take security out there, there's gotta be a leader there or some sort of, um, uh, principle based structure for cybersecurity and risk.

Right. So that needs to happen. Absolutely. Um, you, you would need to have a very, very strong employee experience engagement model. So some kind of like hr, it hybrid. And again, we're not gonna that rabbit hole tonight, but I think that you would need to have that. And then you need, you need governance. Yeah.

Someone needs to watch the watchers and if you just let I it operations run amuck. Or any of these, it, um, adjacent functions run amuck, then your governance is gonna be, I mean, it's gonna be awful. So I think you need, at a minimum, those three things to do, still do decentralized it. How, to your point, how the rest turns out.

Yeah. It's coming by company basis. It's gonna be a, what's gonna work best for right now and we can always evolve to something else, [00:57:00] um, model. But at the very least, I think you need those, those three things in place. Still can't, you can't get outta that. Yep. Um, and we'll talk about the a, the AI influence and decentralized it I think in a little bit.

'cause I wanna, I don't wanna get too far on that whole just right now, right? Yet. Sure. But, um, yeah, so we'll, we'll come, we'll come back to the, the AI part and then, um, in terms of, um, like that, that governance idea, one thing that we had talked about, um, and this was something that you said again in episode 13 of season one, you said good prioritization, governance is not predicated on the assertion that people implementing approved projects know how to implement them.

So that's where project management governance comes in. So, and I like this statement. I kind of plucked it outta the thin air because, [00:58:00] um, in truth it doesn't matter. What kind of governance you have. It's only good if people know how to implement, uh, your project process or Yeah, or your governance process.

If you are the one who's constantly having to shepherd them through, over and over and over again, like now you gotta do a URS and now you gotta do a RFP, now you gotta do, then your governance model is kinda shit. 

Mike Crispin: Yeah. And a lot of times that's you trying to make it happen within your own team. I think that's, yeah, that's hard to do.

And you may already have a program management function in which they operate a certain way and, hold on. Can pause, 

Nate McBride: can you pause that, pause that one a second though, because Yeah. Yeah. You said something interesting. Program management. So let me ask, let, let's talk about this for a moment. In life sciences, there's, there's two different groups.

Yep. Sometimes, sometimes there's program management and there's project management. That's right. [00:59:00] And there's a, there's a clear decision and they get, if you ask the people in program management, they will tell you they're not project management. 

Mike Crispin: They give you, there's a lot of confusion around that. Um, when program management is being tasked at times of one of their programs is to enhance business process, they become project managers.

So if they're, they're away from the programs, which are your, probably your drug programs. Right. A lot of times I think as companies get a little bit bigger, they get tasked with, okay, yeah, we actually need. A way to operate better. And they go to the program teams and they say, you think you could leverage your PMO model or whatever you might have, uh, to help drive a couple of these cross-functional projects.

But I think that's, as you get bigger, and even in, you know, and, and working with my peers and program management, now we do work on some projects together that are, that are more to help enhance and enable the programs. But they all have. But you, you 

Nate McBride: would, you would be the [01:00:00] effectively the default, the de facto pro project manager for technology.

You wouldn't go to program management to manage your technology projects? 

Mike Crispin: Not necessarily. I mean, in terms of the, yeah, I mean if there's, if there's a, you have two people in it, it, they're offering up some project management, which is really just them following your process and helping to organize the meetings and keep people on time and on track and try and manage resources cross-functionally.

That can be a great help. If they're willing to help do that, then it's not as much as that they're doing the implementation, the actual work or setting the expectations around the implementation as much as they are organizing the team to succeed. 

Nate McBride: Lemme rephrase that statement. So it's not their responsibility, but you can lean on them and your organization, you can love them as a resource to help out.

Okay. 

Mike Crispin: Yes, yes, yes. And it's really, uh. Optional. I think I do think you can, you know, if you've got sort of the business [01:01:00] partner, project manager model where you really want that person to also become sort of the SME in, in the space in the certain functional space as well, then that's, that's where you want to empower them to really drive and lead the project.

So they're become the hypercare and they become the support contact in terms of how to dole out resources to the managed services partner or wherever else you're going. Maybe it's internal NY or whatnot, but you want them to remain That I think is the best user experience. How many systems, uh, and technologies have we implemented?

Or you get the, the a pre-sales team and then you get the C implementation team and the hypercare or post support team is just non-existent. You know, if you can internally have that all star at the front and all the way through to the support, you're gonna have a successful implementation and adoption of your system.

Nate McBride: And I think what, what we can infer [01:02:00] from all of that, uh, or what, at least I, I feel like I can infer from all that, is that it's complicated to have successful PM without Oh yeah. Either centralizing under strong leadership. 'cause otherwise Absolutely, you're gonna be leaning on adjacent partners to get that done.

Mike Crispin: It's great when they can really help move things along and keep and hold people accountable as well. I mean, ultimately that's sort of our, our role. But you can't be everywhere all at Watson, if you're running 30 or 40 projects. If you can have someone who's, who's sitting in those meetings and helping say, Hey, you didn't show up to the meeting last week.

Yeah. Can you get, or the follow ups or the minutes or whatever, like to have that. It's just, uh, project management is, it's nice. Is an, 

Nate McBride: it's a people investment that I think it's not at the top of my list for FTEs. And it would require a certain amount of growth and maturity to actually want to spend the money on a project manager.

Mike Crispin: Sure. 

Nate McBride: Um, and you, I think your company has to be at a certain 

Mike Crispin: Yes, 

Nate McBride: yes. Uh, complexity. And you [01:03:00] 

Mike Crispin: right now, I mean probably for you and I, we have to interface with the vendors project manager, which is Yes. Well, that's all gonna be so good. It's a mixed 

Nate McBride: bag. Mixed bag. Man. Don't even get me started. Won't we have, won't.

One vendor, their PM is literally a third grader. Um, 

Mike Crispin: it's literally like, you remember like in, in, in high school or, or you know, junior high school when you, when your, your science teacher, your math teacher was out and they handed out the worksheets. 

Yeah. 

That's what it's like. All right, here's your worksheet.

Everyone in the room, here's your worksheet. You know, make sure you, you, you make sure you do everything. The signup sheet, did you show up to the meeting? And then it's like, well, what did you guys do in the meeting? Oh yeah, we, we just went through the worksheet and, um, everyone finished the worksheet. Look, I drew pictures on 

Nate McBride: my worksheet.

Mike Crispin: I mean, I did a, I get there were, there were a couple projects that, that I was a sponsor of that, that a couple companies in the past. And it's like [01:04:00] what was discussed in the meeting. It's like, well, we just went through the action items. Well, you've went through the action items for the last three weeks.

Have any of them got done? No. We, we just made sure we went through 'em all. It's like, yeah, okay, great. Like, are you driving this project? Or, or, or like, where, what, how can I help? How can I support you and help you make more actions, make this successful? Like, do you need more money? Do you need approval or are people not showing up to the meeting?

I mean, where's the gap? And some of these, you know, kind of project managers you bring in from outside to help, you know, 'cause we had a lot of project work. We're like, uh, you know, uh, you know, I don't know, maybe we have too many projects. It's like, that's not what we hired you for. 

Yeah. 

So that's, I mean, it backfires You try and do too much with even enough resources.

You gotta have the right people and the right plan into your initial point. You've gotta have the right project governance structure in place that works. Sure. That's tried [01:05:00] and true. And it, it's, it's gonna fall down the first few times. You just gotta find one that works for your company and your culture because.

It's, it's not, it's not easy to do 

Nate McBride: with. It can also get dicey too, when, if you don't have the absolute topnotch project management in place and governance around it, even if you have governance within it, uh, uh, project management governance, which typically also involves business analysts. They're work generators for it.

They're not work. Um, yeah. 

Mike Crispin: They're advocates leaders for the, for the functions. For the functions. 

Nate McBride: Yep. So they're bringing work in, which isn't always necessarily the best thing, especially when people are already sort of stretched to the limits. But I, 

Mike Crispin: but to your, your point around project go project management governance, it doesn't have to be that difficult.

I mean, it can be, Hey, we got one central master list of projects. This is the data we need to capture. This is what we need to review every week. This is the schedule. Here's [01:06:00] two or three sprints we're gonna go through. Does everyone agree? Yeah. Everyone agrees. But I mean, we had, you know, back, you know, I'm sure we're not, I think I, I would guess that you and I are not waterfall, um, fans at this point at all.

And I'm certainly not. But the amount of paperwork and rigor order to having, um, 

Nate McBride: yeah. Don't even get me started. 

Mike Crispin: Waterfall is just gonna turn people off. To getting things done. It might protect you in the Yeah, in the, in the, in the, you know, the CYA and the, when things don't go right, that you've done all the extra work, but the end of the day there's so much extra work just to, for it's really risk reduction that you'll probably never face, you know?

Nate McBride: Totally. So, um, all right. So switching gears off of so decentralized it, not much has changed. We'll come back to the AI part in a bit. Um, I did wanna talk about one more part of it though, which was the employee experience idea, [01:07:00] but I don't wanna go down the, the human development hole, but I did wanna ask, so I wanna ask a question about employee experience in general.

Um, we, you know, you and I have the same pressures, which are, when we hiring somebody to attract them to our company, like, we have to figure out sort of what's the, what's the hook, right? Mm-hmm. And when someone says every single company sounds exactly the same, you know, the same benefits and the same, uh, paid time off structures and all those other things, the differentiator, I think comes down to word of mouth and employee experience 'cause such tight industry.

And so, um, just a bit general question, but in your opinion, um, has, has anything changed? Um, in terms of how companies that you're aware of, including your own, and this is an indicting statement, you don't have to say anything about it [01:08:00] towards cardian, but how, like, just in general, how have, have any, has anybody changed the employee experience model?

Is it something that even needs to be farmed out to its own function or is it so just opaque that and it doesn't seem like anyone cares so much about it? 

Mike Crispin: I, I think a lot of people in a lot of functions and leadership care about it. I, I think it's just a hard thing to consistently stay focused on in the respect that some employees have all different requirements and things that they think are important when they come into a company.

And each function, I think, needs to do their part to orient people with the company, whether that be in the interview process, you know, you're in your first two, three rounds of interviews, you really wanna be transparent about how great the company is and the challenges the company has. [01:09:00] Right. And it's like things that you, so some people feed off of challenge, like a lot of times try to butter up and sugarcoat things For sure.

Yeah. Like a lot of people you interview, they're, they're gonna. You're going to be straight about what the challenges are gonna be and maybe they can help, but that's, it's all, that's what I'm trying, I guess the point I'm trying to get to is it really depends on the person and it, it gets difficult to, um, to kind of put the, a consistent messaging together.

I'd say the most important thing is orientation. Once you're in, have an orientation, it'd be, it's great to have all the functions participate as much as possible. Yeah. But if you can work with within your control to all to ear in the early days, try and build as much trust right out of the gate as possible.

That's super important. But you, it's, it's, um, some, some people won't think it's a good experience, uh, in terms of things that you and I, Nate can control or ahead of it can control. Yeah. Some people won't think [01:10:00] they're a good, there's a good employee experience. Unless you're on all Microsoft products, period.

They're not gonna be happy. They're not gonna be happy unless you're on SharePoint. Unless you're using Outlook. Unless you're using this, this thing they've used for a long time. It and, 'cause they've gotta get a job done and they don't wanna change. They got this very important job they gotta come in and do, which I'm not knocking that in a new way, but they're used to a certain set of tools and if you're in there telling them, well we've got this cool new thing and everyone.

Is excited about it. And, and some people might kinda go, oh my gosh, like what did I do? Yeah, I, I got to relearn all this stuff. Uh, and then there are others that come in and they feel like they're part of something new. You're put, you got a, you've got the, an AI capability that people are excited about or you're making it so they can choose their own laptop or you're make it, they have, they have access to tools and training and they can build their personal skill sets and everything.

Different people will be excited by different things. Yeah. In terms of the realm of things that you and I or leadership and the IT [01:11:00] organization can control. So, sure. I think there's a lot of thought about it and I think there's a lot of care that people want to put into it, but unfortunately, I'm not sure, like sometimes it rises to the top.

It may also be very much based on the hiring plan of an organization. If they're gonna hire, if there's an inflection point or milestone coming up for the company and they need these key people, they're gonna spend a lot of time on just that employee experience to retain the people who are there and to bring in the top players that they need for that to reach that milestone.

But if the company's only hiring a few people in the year, or it's, um, backfills or whatnot, maybe there isn't as much of a focus. And again, just speaking from 

Trance Bot: assumptions, no, this is wise, 

Mike Crispin: just speaking from assumptions, not from really anywhere in terms of where I stand today. Yeah. Which, which is more, I'm always trying to focus on that and I, I wanna like where I work, but I also want people who come in to [01:12:00] like where they work and if, uh, my team can be a part of that, that's, that's a huge success in my book.

Nate McBride: So, uh, dig everything that you're saying, but it sounds like you're making the case for employee experience being important enough 

Mike Crispin: Oh, yeah. 

Nate McBride: To potentially, to potentially warrant its own, um, lens, its own unique lens. Yeah, 

Mike Crispin: it, I think it does. And I think that in some respects has changed post COVID, right?

So it's, yeah. Yeah. We do see still a large amount of companies that are just, especially new companies that are gonna be remote by default. But we also see, I I also hear, I should say, just of many companies trying to get back, uh, in the office and having a, a lot of incentives to bring people or have people move back closer to the office.

Yeah. 

Nate McBride: Yeah. 

Mike Crispin: So I think there's [01:13:00] different, different schools of thought on that. I think they're both good in different ways. Um, but, you know, you and I talked about identity last week and we're getting to a weird place here, and we're gonna, it's gonna continue to get weird or we don't have to go back to it, but.

People being able to see each other in person again, there could be a lot of de-risking benefits to that becoming more of a reality. Again, not just, not just from a cybersecurity perspective, but just from a, um, personal health perspective. Um, so That's right, and, and, and especially young professionals. So 

Nate McBride: let me, let me, let me uh, phrase this, uh, one of, let me give you one more angle to think about and then we'll move on.

Um, you just, 'cause you just mentioned something that's very, very near and dear to me. Let's suppose that, you know, we, we start, like companies start bringing more and more people in. Yeah. And the original thesis was that employee experience would be [01:14:00] moved out of it into a central function. Now that thesis might have been a little bit misguided originally because the, it all, it infers that it owned employee experience.

We only own a portion of it piece, just a piece. A piece of it. So, so in light of what you just said, let me rephrase the question and see if, see how you think about it this way. Okay. What if the employee experience piece. Just was centralized out of it along with other employee experience, people into a central function.

You focused on making sure that those standards were met, but you didn't own the employee experience, somebody else did. Do you want to continue to own the IT portion of the employee experience or would you be okay in a decentralized model of moving that functionality out? So, 

Mike Crispin: I don't know, maybe I've changed course a little bit on this.

I think the entire g [01:15:00] and a and risk functions of the company own a piece of employee experience collectively. Sure, sure they do. I think that, I think that it's, sure, I do think people think of, uh, the people team or the human resources team, um, as owning that piece, but it's almost like the same thing when we say company culture.

Company culture is owned by the whole company and almost organically, hopefully more a grassroots movement for a lot of companies than it is being the people at the top trying to dictate the culture. I hope that's how it's gonna turn out and how it is for most companies, I think in some 

Nate McBride: cases or 

Mike Crispin: more progressive.

But that might also, that might be also true of employee experience is ultimately the sa on a d slightly different wavelength is that, uh, your experience is kind of your culture. So if you take it in the. And, and respect that yes, there's technical, uh, operations and benefits and, you know, actual work that needs to be done that's mandatory for each state and, [01:16:00] you know, each business legally.

Um, that there's, it's not just kind of building the framework and having to execute it. It's also lot of the tactics as well that certain functions own. But I, I think it's a shared, it could be someone continuing or a group. If, if, if we're talking about kind of separating the group, which I can understand as helping to, just like you have a cult, cross-functional culture team, like someone who's helping to remind the organization of how important is that we stay engaged with employee experience and that it is important with everything else that's happening.

Yeah. Um, that maybe it's more of a, a, a collective within the company that helps to, that helps to just keep things visible and keep them at the top. Because sometimes onboarding, even onboarding and offboarding processes, that can fall to the bottom of the priority list because other things happen or come up.

Yeah. Or maybe this hiring plans have changed and it's like, oh, we wanna spend all this time on this, but employee experience is [01:17:00] so much more than that. So it's, so maybe, maybe 

Nate McBride: it's decent. Well, no, but in decentralized model, the point was to distill it down to its operational core. So in that case, if we just check out ex from Decentralized IT model, it would either have to stay in the operational core, uh, elements, or it would have to be performed out.

I mean That's right. That's what we're talking about. 

Mike Crispin: And I think the traditional, traditional model, like we, you see employee experience as part of HR is part of like there's an employee experience lead, uh, that's in place, you know, you Yeah, yeah. Have employee experience sometimes. But now I think, I think because of the, because it really is a shared, uh, I think experience that functions are trying to put forth, and maybe this is part of it, is that maybe finance and legal and some IT departments don't think of themselves as being big contributors to employee experience.

They're not, [01:18:00] it's, it's not on the top of their mind. It's more like, it's kind of a given. Yeah. They can log into their computer and everything works, and I'm not getting a lot of calls. Great. Um, you know, people are getting their paycheck and they, they can do their budget and um, you know, they understand the spend policy and the signing limits and everything else.

Great. You know, that's, we got all those things done. It's the soft pieces in between that, um, maybe aren't at the top of the mind, and the collective could help bring some of those out. 

Nate McBride: Well. And, and, and truthfully, there's another element to all of this, which we never talked about in that first episode.

Um, and you and I have only briefly touched on this a few times in the past few years, but it is the fact that if it is going to decentralize 

Mike Crispin: mm-hmm. 

Nate McBride: Can it decentralize by itself in a g and a function, or should all GNA functions decentralize? Because if I just say, okay, you know, legal, you stay as legal, you'll have IP and you'll [01:19:00] have, uh, contracts and you'll have, and then HR you'll stay with, you know, benefits all things.

If everyone's gonna stay as cohesive units with all their stuff, and only it is gonna decentralize, that doesn't make a lot of sense to me because we're the Oh, agree. Why are we doing it then? Yeah. Yeah. So, so I think, uh, the bigger, the bigger talk is that if, if one GA function is gonna decentralize, it would be optimal potentially.

Mm-hmm. If all GA functions in the same company. Decentralized. 

Mike Crispin: Yeah. Yes. I, I think, um, and that to, to that point, the hybrid or extended model or dual reporting or maybe more to matrix model that you mentioned. Yeah. That was the 

Nate McBride: dotted line reporting model. 

Mike Crispin: Exactly. Is that, you know, we do, I know, I, I know at uh, last company, you know, there are HR business partners for each group of functions.

There were IT business partners. They sat out with the business and there's always a question of [01:20:00] whether or not what their reporting relationship should be. And then from a a legal perspective, we had sort of dedicated legal partners that worked with certain functions directly and they reported back into legal.

So this, that's sort of the distributed, maybe they were reporting into the head, but they, like we talked about business analysts, were advocating for their customer. If you want to call us that or call them, call, call up their people they're working for, or advocating for, I should say, um, that, that they, they were ultimately really part of your group if they were doing a good job, you know, and we would always talk about the IT business partners.

If they're not being invited to their customers meetings, then they're not really part of the team. So like, I think that's been a struggle. And, and businesses are slightly sl larger, maybe more medium sized businesses. They reported into the functions directly. Um, and you could have your own [01:21:00] liaison, if you will, that would pretty much have either a dotted line or would report right into some within the group.

So, I don't know, I think everyone's got a different structure, but when you start adding, and I know we're gonna talk about this time, but you start adding, um, non-human elements to this, that's when it, it, it starts as that the glue, well, we'll talk about next in the future, but I'll just say that's, I think that's one of the things that could change the.

That sort of move the needle a little bit to change the direction. 

Nate McBride: Um, you know, it's, uh, you don't know this maybe I, maybe I told you I listen to a lot of other podcasts and one of them is the Jocko podcast, the Jocko Willink, and he's a former Navy Seal, um, very, very influential sort of, uh, leadership type.

Very, very big on the leadership style. Um mm-hmm. It has always has a lot of great insights. And I listen to a lot of these podcasts and they have former Marines and Navy Seals on there talking about sort of their leadership and their time spent in Iraq and Afghanistan, [01:22:00] um, in Vietnam. And, you know, there's a, there's a, an idea that keeps coming up every time they talk about, um, what was happening in Ramadi.

And a lot of these sort of smaller groups would split off. You'd have a medic, you'd have a sniper. Yeah. You'd have, um, uh, some, uh, actual like gunners. You'd have maybe a helicopter unit or a Bradley unit attached to it. She'd have people from all their respective disciplines going together to go solve a problem.

And I think about that in the corporate world, like why do we isolate ourselves in functions when, if I need to solve a problem, what, what we do, call them our project teams, but we come together and we split back apart. Imagine if you had these project teams. And a corporation that never left each other, that they were banded together by a single purpose.

And then mm-hmm. Um, for instance, you'd have all your people that were focused on risk in a function called risk. I know it sounds really, really [01:23:00] radical, right? But take, take your legal person, your CISO and whoever else gives a shit about risk. You know, your, your chief accountant, put 'em in a function, have the seniormost person lead it, and there you go.

Boom. You have a new function. Right? Um, and it would certainly be able to attach itself to units or these functions as they went along. Um, maybe either permanently or at least for longer terms to provide support. So, anyway, interesting. Interesting idea. Potentially, I thought about that one a little bit.

Like how would I, how would I reshake up that model? And there's a lot of good, there's a lot of good teams I would put together, like on day one that would absolutely kill it. If they, if you put these people together in a room and let them work together for like six weeks, holy shit. Uh, what they would get done is, um, amazing.

We didn't talk too much about fully centralized it 'cause it's a super fricking boring and nothing's ever [01:24:00] gonna change. But in all seriousness, like, um, those companies that are still doing fully centralized, like matrix to it, you know, bottom to top. Yeah. Um, that model. It makes sense when you have, uh, a lot of, in a lot of different verticals reporting into an IT leader for sure.

Yep. You have development, security, project management experience, bigger companies, infrastructure, cloud ops, um, so you want all those heads reporting into a single leader directly. 'cause there's a lot of tr dynamic operations and activity. Sure. I don't think that's changed at all. I think that's pretty much exactly the same.

And it's certain kind of people that are gonna build those structures 'cause that's what they were trained to do. 

Mike Crispin: Completely agree. That's not not knocking them at all. 

Nate McBride: It's efficient. It's, yep. It's a, it's a function that does well. Um, it only time it really ever doesn't work is when you don't have fast decision making up and down.

Like if you don't give your [01:25:00] leaders enough autonomy to go out and make their own decisions and they always have to come back to you or the people have to come back to them, then it can slow it down. The business can get a little bit miffed about that. 

Mike Crispin: Yeah. Gotta gotta let them operate and at times you gotta remind them that you hired them to run this thing.

Yeah. This is their, their role and this is their decision to make. You're supportive. That's why you hired them. And it pretty much, you guys get to a bigger, you, you've hired many big departments. It's like once you get to this point, it's like. I hired you, 'cause you, you know this stuff. I'm not the person who knows this area.

You're the, you're the genius. So 

Nate McBride: expert. Here's like the area, here's the, here's the guard rails, here's the area of operations, and now go, yeah, go for it. This is, this is why I hired 

Mike Crispin: you. I can't wait 

Nate McBride: to see what's going on. That, that brings back a high degree of speed and efficiency. Decentralized it. Yes.

When you're, you're giving autonomy and [01:26:00] enabling all of your direct reports and they're doing the same to theirs to move fast. Um, yep. We, we do what they need to do. So I was back at TKT and Greg Perry, we call the 70% principle, which is, uh, once you get 70% of the information, that's good enough. Like you don't need a hundred percent.

Um, you don't, I love that guy, man. He taught me the two by two matrix and I still use it to this day. It's a genius decision making tool. Um, Feder and, and we talked about Matrix IT and the federated model. Oh yeah. Um. I think we all, we got really hung up on Federated, and it's, see in, in episode 13 too. The, the only thing we all, I think agreed on with Federated was that, um, it doesn't work when the hub becomes a bottleneck.

Otherwise, we were kind of all over the map back then. But like in terms of federated, where you actually [01:27:00] have shared services, you know what, what it should actually legitimately be, in my opinion, which is like shared services when you have shared services, so you have a cybersecurity division and, uh, operations division and governance division, kind of like decentralized it, but literally these are not IT functions that are separate.

They're literally their own, their own divisions. Um, and then operational, IT is distributed throughout all of them. So like each one of them may have a cloud administrator in their group. Each one of them may have a experienced person in their group. So it's not like experience is a group experience is distributed among all these other groups.

So literally every one of these federated groups almost looks identical, but they all have sort certain unique strengths. Um, I don't, I don't know of a, of a certain example today that is using, uh, pure federated it. I don't either. Yeah. Um, nope. I think, I [01:28:00] think this one might, might have gone the way the Dodo, the judo.

Um,

so, all right, I wanna get to the AI parts because Okay. I think that's where our conversation's gonna take a little bit of a left turn, so, sure. And we can, we can start with, um, AI and, and decentralization. And when I say ai, I'm talking about Gen ai, of course. Um, okay. So if we think about all the crazy shit that's going on right now in the world, all the jobs being eliminated for AI people, and how we all know that that's ludicrous.

Um, I feel like, I feel like there's a, there's a, like a tension in both directions for and against decentralization. So ai, um, gen AI [01:29:00] rather for, for decentralization, um, means that, you mentioned this earlier in the podcast here, but, um, business units are getting smarter and people are getting more and more savvy so they can now leverage AI to make their own technological decisions.

They're not. Uh, there, there was a great insight I heard the other day, and I'm trying to remember who it was. Um, my name Malcolm Gladwell who said that, you know, the difference between people today and 20 years ago is that a leader kind of knew more than anybody else. Mm-hmm. Like they had special information that they knew in addition to being a good leader, they, they only knew certain things that nobody else knew.

But nowadays you just like one agent prompt away from knowing as much as your boss, uh, at any given time. So leadership is like a special sauce skill now. Uh, more so than what, you know, 'cause now you have to assume that everyone knows what you know. [01:30:00] So with that, I think the case for decentralization is that the writing is on the wall.

Every person can literally, within a few well-crafted prompts, know as much about a thing as I do. 

Mike Crispin: Mm-hmm. 

Nate McBride: Yeah. Which, why am I me? I, I don't mind that at all. So it reduces, well, reduces dependency. Exactly. Absolutely. Um, agreed right outta the gate. And, um, I don't know about like speed and efficiency. I mean, those are always, uh, marginally relative gains when using gen AI to begin with.

Then to go out and say that, yes, by doing this ourselves, we've saved this much time and money. Um, 'cause it also makes you question like, how shitty were your processes to begin with. Mm-hmm. But, um. Then the question becomes, okay, like, let's, we'll pause almost in a, in a second, but the question is, [01:31:00] if that's the case and you really want to extend this thinking out, does that mean that every employee, like again, you take ex out, you take, uh, and that becomes its own thing and governance out and security out and you have IT operations, could it be that every employee has, uh, Claude be their IT support?

Mike Crispin: I even, yes, and even more so I think where run some risk when we talk about governance and all these unes, unsexy, de-risking things is that it in some respects and probably other GNA functions are looked at as bottlenecks and with these tools, it's gonna just exasperate it even more. Yeah. An employee can build an automation pretty easily.

They can build a data set, they can run reports, they can even [01:32:00] potentially build integrations. We've talked about that with Zapier and the, there's a number of things that they can build. Websites, applications, dashboards, the stuff's all gonna get easier and easier to do, and. Largely it organizations probably 'cause they, maybe they understand the risks and cybersecurity professionals are gonna keep saying, no, no, no, no, no, no.

Follow the rules that we have. 

Nate McBride: Yeah. 

Mike Crispin: And, uh, I don't know how all that's gonna work out. Well, you gotta, you 

Nate McBride: gotta admit though, it is a bit of a governance nightmare. I mean, you are unleashing Absolutely, absolutely. Governance 

Mike Crispin: tell. I think that's why we may need to think about how we can take a little more risk, but also make the organization aware, take a similar approach to what we, we continue to invest and build in cybersecurity.

And that's just awareness. Less about governance and more about a softer side, which is awareness and employee agreements. Yeah. And everything in between that is, [01:33:00] is, is, is, is, um, you know, is is extra work. So as I'm just saying, as these tools emerge, I don't think it is true today, but as time goes on, uh, you can, there's gonna be a lot of things that employees can do that we will have to trust that they won't do.

And really the only things that protect us from those type of things today is. Your employee agreement, your policies, it's not gonna be technical guardrails or technical protections. It's gonna be you promise not to do this, right? Yes, I do. Well, and that's because they're gonna be able to, to do it. Not only that, but you're gonna, 

Nate McBride: you're gonna end up with inconsistent approaches across the org.

So, um, yeah. So, so Mike does it this way and Nate does it that way, and they're inconsistent. Exactly. And then Exactly. Obviously the security and compliance gaps. I mean, fucking hell. But, 

Mike Crispin: but, and Nate, it just takes a, a, a few companies that are moving very quickly to fall [01:34:00] down and see some real impact for there to be, um, I know we're gonna talk about this maybe next week, but this is where the verification economy enters.

Its its head, right? So is, by the way, it's not about, that's not 

Nate McBride: for 

Mike Crispin: episode four next week. It's not about governance anymore. It's about verification. How do we verify the automations that employees make work correctly? How do we verify that the data is good or that these reports are accurate? How do we verify that the answers that, uh, we have gone through are able to be traced back?

Um, and similar to what a validation, uh, exercise would do in a, in a performance, quality or PQ scenario, like I just have to ask a question. Think there's the whole verification department. Hate. 

Nate McBride: Yeah. I have to ask the question that we, of governance, if we have to get to that fucking level of having a verification department, maybe we're going down the wrong path, uh, by even considering this.

Yeah. But, but I'm just, you know, to give the's a pragmatic or really ful [01:35:00] viewpoint. Well, 

Mike Crispin: I think, I think if you had the verification model today, you'd have a lot of upset, offended people For sure. And for AI is not, AI is not gonna get offended if you check its work unless it's out to view on life support or something.

But, but I think that that's, that's, I think that's the difference is that there are probably some areas in which we'd like to verify work that's been done, but there's an element of, obviously you have trust in your employees and your people. We probably have more trust in this AI thing, which is gonna be crazy, right?

Yeah. I mean, that's what we've been talking about. So there's gonna be assumption that it's correct. And I think that's 

where this, that it's gonna have to be, that's also a problem. 

Yes. Yes. Right. So I think that's where, as there are lessons learned, the need to verify certain, certain aspects and certain decisions.

You say the human, what do they call it? Human, um, what is this? Is the, the in the loop tagline. In the loop. In the loop, right? Yeah. [01:36:00] Human in the loop. As long as that is important and people understand that, which I think most people right now are still hugely for that, and I, I'm certainly one of them, is that the, the day that that changes and it's.

Wow, I haven't seen this thing make a mistake in three years or, or you know, in, in six weeks even. Uh, this thing's never screwed up anything. Uh, let's just give it everything. Yeah. Um, I think that's not gonna be a unpopular decision someday, so I dunno. It's gonna Interesting. Well, I think it's very interesting.

Nate McBride: Well, this is, remember we're, we're making the case for decentralization, so That's right. You know, the reality check is, yeah, it might lower the floor for entry, but it doesn't remove the need for expertise. You still need that human in the loop. Right? Right. You do. And that, and that false confidence only means that you get bigger disasters faster.

You don't like That's right. And you don't know how they 

Mike Crispin: happened, you, how they [01:37:00] happened, you know, why they happened. Other than that you, maybe you trusted something you shouldn't had. But I do think the decentralization for the, the maybe some of the low hanging thing, like, let's take chat GPT right now.

You know, if you don't have an AI tool of some sort, or you haven't been doing training on AI or anything, or raising any awareness, chances are everyone in your company's in chat GPT when they get home. You know, like, what do, what do you expect 'em to do? Um, and I mean, that's certainly one of the things that, you know, as we work on our strategy is to make sure that we're giving people a, an avenue in which to understand the policies and the rules and, but even I think that.

As these tools get better and better and more accessible, if someone can skip the line and get ahead and get something done faster and get a promotion for it, they're gonna use the tool that you don't want 'em to use. I, I just think that's the [01:38:00] reality of some of these, of the way the world's going. And you, that's, that's decentralized 1 0 1 is I don't need someone to gimme the rules.

I don't need someone to help me stick to the, to the, to the guidebook. I just raised this much money. I just fixed this many, this many problems. I just signed up this new client. I just did, I had this business impact and I didn't listen to Nate or Mike. Um, so that, that some people will be willing to look the other way until something really bad happens.

Nate McBride: Yeah. And then, then, then the problem is people that have invested there, um, sort of position and stake in that it's going to work, aren't going to admit when something bad happened, which is another problem. But, um, 

Mike Crispin: but I think your point around decentralization, the more things you can use to have, not have to ask questions, [01:39:00] uh, or what I said being like, the easier it gets to accomplish certain things that the more, if an IT department is constantly saying no, uh.

Then they, it's just got, it's just worse than it even is now. Some of the, the things that we need to put in place and that we, you know, we try and explain why they're important and they de-risk or make us more productive. Um, as these, as the tools get more accessible and they start to work better. Just you, you need to en engage and embrace more than stop and slow down.

Nate McBride: Oh, it's, uh, it's, sorry. No, it's, it's b it's, everything you said is correct. It's fucking baffling to, to, to think about that. We're at a point where we're, we're conditioning ourselves pretty routinely to consider what it would be like to not have a human do this thing that's so important for us. Mm-hmm.

And then trust that the, um, it's gonna work. I mean [01:40:00] yeah. A decentralized it model. You build an AI thing that can do a thing. Who fixes it when it breaks And, um, well, right there you have your first problem, uh, among all the others that come up. I mean, it's just, uh, I hope AI can fix it. Yeah. Yeah. Me too. Me too, Mike.

Um, so, alright, let's, let's, let's, uh, as they call, let's as they say, let's put a pin in that and come and come back. Let's double click into, uh, why the, the pros of AI for centralization. So that, that was the pro, that was the discussion about decentralization. Let's, let's double click into to why it's great for centralization.

Um, obviously all the things that we do today. So you're getting better compliance control in theory. 

Mike Crispin: Mm-hmm. Mm-hmm. 

Nate McBride: Um, [01:41:00] you are getting the cost management centralized. 

Mike Crispin: Yep. 

Nate McBride: You're getting probably authentication centralized, hopefully. Yes, sir. Um, you're getting the governance pieces centralized. So why do we use this and not that?

Um, you're sort of taking control over properly, putting guardrails around the use of gen AI in a company under centralized IT model. 'cause you, 'cause everyone in it is onboard. Yes, 

Mike Crispin: yes. Yes. And, and, and, uh, go. I'm sorry, go ahead. No, no, keep, go ahead. I'm thinking I was, I I, I was just gonna say one, one other area is it's just the collective business principles of the company.

As you're putting in ai, you're certainly probably thinking about that employee experience aspect that we talked about earlier. [01:42:00] And if everyone's using a different AI model and a different tool, the collective culture of the organization and what that model knows about your company and how it's going to help foster your culture, uh, you want that to be unified.

So you don't, you want to eliminate what biases you can, you want to have the ability to have it understand your company norms and company culture, and you want that to be a unified position. So you, that needs to be centralized in, uh, in a way that we understand the models that we've brought in that align with the, the corporate values of the company, uh, with the ability for us to understand how decisions are made and also to, and control 

Nate McBride: control dependencies.

Mike Crispin: Yes. And, and to push back when someone pushes too hard into the AI to make a decision. So that's, that's a really important thing I think as we move forward, that if you've got a, if you've got [01:43:00] a, what do we call it? A what, what is it? The a yes man or a yes woman, right? Uh, is if you've got a yes. Ai, which has been a big criticism of the, the, the previous GPT and Gemini models is you need AI to, to push back and say.

You shouldn't be asking me this, right? Or this doesn't make sense for your company. You know, you probably wanna go ask your boss, you know, or whatever, you know, whatever you, however you want, it should, it should be set up so more that there's a centralized function, it's gonna help to mold that, that model, and that you're, you're using the, uh, similar to the same, same rules we have with applications that we don't want a sprawl of inconsistent responses and data.

So that's, that brings the centralized function back full circle to where a a a lighter touch sort of verification or soft, more of a soft [01:44:00] employee experience stance is less than a governance stance and more of a, we're gonna foster the model. We're gonna, we're gonna expose the model, we're gonna have a, a, um, a usage, you know, obviously an acceptable use policy that's based on enabling you, based on making you feel like you're part of something special.

And not just will the people at the company do that, but the tools now have a life to them. They now have a personality to them that are gonna help you feel and make sure you are part of something. And I think that comes back to the employee experience full circle, is that the best models for your company down the line are gonna be the ones that you feel you can, you can trust with data, but you can trust with your business problems with your personal development.

But that there's still humans in the loop to help foster that, that knowledge and that conscience within your company. And that's why I think it needs that, in that respect, needs to be centralized. And it's less about [01:45:00] governance and more about how do we, how do we build this sort of digital culture within these tools that we're inevitably going to have to compete with other companies and with hiring practices.

So I, yeah, I think there's, it's exciting to think of it in that lens where it's not as much, we're putting rules in place that are gonna make you look at it. Oh my God, we thought it was bad before. Now they telling us we can't do this, we can't do that. Now. It's more about we want this thing to em, we want to embrace, to embrace you with the things, the problems that you have and help you get your work done as opposed to, um, you can't use that.

It's gonna be more like, you may not be able to use that, but you're gonna love this. 

Nate McBride: Yeah. 

Mike Crispin: And so it's, that's gotta be, I think it has to be centralized in a shared responsibility in some respects between some of the core experience led functions in the company. 

Nate McBride: It's, uh. Man, there's a lot of things I wanna say.

First of all, the way, the same way that you mentioned earlier, like, you want some, someone [01:46:00] wants to come in, they just wanna use a tool. Mm-hmm. Like, I just wanna use SharePoint. Why the fuck you gotta gimme box and Slack and all this stuff? I don't want that. I want SharePoint. Right? Yeah, sure. Yeah. Well, you, you can easily visualize two years from now, but I don't, I don't wanna use it.

Claude. I want to use GPT, I want to use Atlas. Like that's what I'm comfortable in. And you're like, well, sorry, you have to learn a new thing. Um, I can easily see that becoming a reality. Right. Um, 

Mike Crispin: I think if done right, donate that. It doesn't, they'll, they'll, they will interface with whichever model the company or whichever set of models or tools the company uses.

Yeah. I that they shouldn't be able to tell. 

Nate McBride: You're talking about the sort of the homogeneous prompt, right? The prompt that just 

Mike Crispin: the headless prompt I'm talking about. I'm talking about, uh. Exposing the API and and Yeah. Having your own ui. Right. Exposing prompt. I mean, they won't know 

Nate McBride: that. And, and all the, their only knowledge will need to be in how to write prompts.

They will have no bias towards the [01:47:00] platform. I get that part. And that's likely a, it's hard. It might be hard to 

Mike Crispin: do that. Yeah. 

Nate McBride: But unless you're at a company that says, you know, these companies that are demanding employees have a 2026 goal, that, that at least one goal has something to do with ai. Right.

This is fucking absurd. But companies are doing it and then they ha you have to, you have to make 10 GPTs next year. Right. Uh, or whatever. I don't Sure. People have lost their minds. Anyway. The central, the central, the, the centralization of it. I think like everything that you just said, it lends itself to that.

Hey, we have selected Claude as our central engine for the company and LLM. Therefore, that's that. Like, that's who we're, who we've chosen for. Right now, it's subject to change like all platforms, but that's our central platform for the company. Done. Right. Full stop. That's the benefit of centralization.

And that makes your governance around that, I think, significantly easier. You're not having to govern multiple engines, multiple platforms, [01:48:00] multiple methods. You're streamlining everybody into a single set of guardrails. Yep. I mean, as you would with any centralized it prompt. I mean, it goes back to. Email and file sharing, it's, it's gonna be this.

If it's not, it is already the same thing. It's a SaaS application. It's no different than anything else. So I think the upside for centralized it is that you can in fact just go ahead and centralize on it. But there's a lot of, I think, downsides to this too, which is of course you're picking a central model and people want to go play with shit.

And so I don't know which shadow it is. Worse the shadow IT and decentralized it or the shadow it and centralized it. And I love my shadow IT people I do, but mm-hmm. Generally I feel like it's more likely that there's gonna be more and worse shadow it in a centralized model because there are so many barriers to innovation unless [01:49:00] centralized it goes out of its way to provide like sort of this innovation model, innovation hub.

Um, I mean, thinking about autonomy for a second, which of the two models gives it the greatest sense of autonomy? Centralized it. Centralized it provides the best sense of, um, solving it autonomy from zeros in one decisions. No, no question about it, but. And centralized. It does not have to be itself a zero model, but it does mean that, um, you've eliminated a lot of the variability that comes with decentralized it.

So including what you can and can't do with ai. Um, so yeah, I was, I was trying to come up with a thought around like 20 27, 20 28, what will we do our prompts [01:50:00] in, um, by then, and I, I said headless before, but, and you know, I think you and I are talking about the same thing, but I think it's gonna be even different.

It's gonna be different than that. Like there'll still be the Anthropics and these other companies around doing God knows what at the time, building missiles or killing countries or whatever. But they'll also still have, uh, a ui, an interface. They'll still be mm-hmm. Needing to make money. And the, the, the wage, the war that's gonna be waged against open source internal corporate token, uh, sorry, token based LLM development is gonna be massive.

Mike Crispin: Mm-hmm. Mm-hmm. 

Nate McBride: Uh, they're gonna wage all out war on people like you and me. They don't want us to build our own open source, um, agents and lms, 

Mike Crispin: but we're going to Oh, yeah. I think, I think there's no doubt about it. I think it's, I think in some respects they're betting on. On that because we'll probably use their tools to build them, [01:51:00] potentially.

That, or they, they assume that anyway. So

Nate McBride: I think back to decentralization like ai, which is to say gen ai, I keep saying ai damnit, gen AI makes decentralization very technically feasible. Um, 'cause business units can basically do more independently, they just need to be let go. To do that, someone needs to sort of take the leash off. Um, with ai, gen ai, you, you don't need it centralized for many tasks in the business.

Mm-hmm. And, um, it enables the business to grab onto that fricking recu like recurring headline, faster, more efficient, better, you know, all these sort of false promises to keep coming up. But maybe those in fact are achievable when everyone in your company is [01:52:00] allowed to do their own thing. Um, basically, but it, it also makes centralized it necessary.

In the same vein. So like, just as it makes decentralized, it technically feasible, I think it makes centralized it necessary because it's the, the risks are too high. You, if you don't have centralized costs, they could be through the roof. Especially if you're having people that are building models on like bedrock and also buying enterprise licenses for themselves.

And then the compliance is just, I mean, I don't even know where to begin, where you would begin on the compliance management for decentralized, um, AI effectively. So I think what will happen is, again, playing this forward knob a little bit, while we may stay with a centralized I it model, I think we might end up going [01:53:00] towards like a federated ai, AI model.

Mm-hmm. Um, right now, like I've restricted my company down to two platforms and maybe that, that I have to let a particular group do a different thing away from that model. And then they are required to self-manage it based on some other guardrail. But I'm, uh, I'm sort of, uh, I'm giving a special exception to the centralized rule, so it remains centralized, but I'm decentralizing the AI components of it if you, if you get me.

Sure. 

Mike Crispin: Makes sense. Makes sense. Um. How do you feel about just the overall data security of these platforms today? I mean, they, they follow a lot of the same SaaS based encryption rules as box and others, like, is it any different really? 

Nate McBride: We just literally, we literally just had this conversation today. I've been training this, I've been training a company, um.

On Gen ai, it's a, uh, [01:54:00] consult, a consulting firm. And they brought me on board to do this work for them. And we talked about this today about what, what can you put into a prompt, basically, right? And I was like, well, um, prompts don't ingest what you write. If you did that, the hackers would have a field day of just poisoning the LLMs.

Instead, what they do is they, of course, mine and farm public documentation, they will sometimes the, the less, uh, ethical, uh, platforms will train their models off of you. But on an enterprise license level, if you go by the terms of service, they are in fact not doing anything. And I said, but here's the corollary.

If you're using an exchange and using Outlook mail right now, or using Teams or Slack or using anything for inside your company to transfer data from one place to another, guess what? It's the same fucking thing. 

Mike Crispin: That's right. That's I said that a couple days ago myself is Yeah, you, you're, this is no different, that it's just a system with [01:55:00] storage.

It's difference 

Nate McBride: than you putting a, the, the confidential recipe inside a box. Exactly. 

Mike Crispin: Exactly. Exactly. So, and, and there's only, I believe there's only one or two of the vendors who have a truly zero trust model or that you can keep your own keys. Um, yeah. So I mean, I, I, I think we're taking the same level of risk as we would.

Using any other sets based application, however, 

Nate McBride: oh, that, that being said, I would, and I, this is what I ha, this is what I consulted them to do. I said, think about whether or not your clients would, would approve of you using their information in a prompt. Now you may know this, but they may not. So it's like simple redaction.

Find replace everywhere that you see Mike Crispin replace with Santa Claus. Everywhere you see the name of Mike Christmas Company replace the widgets incorporated. So the outcome will still be the same. Then you can reverse redact in the final output. 

Mike Crispin: Mm-hmm. I agree with you. But if they [01:56:00] did, if they did leave that stuff in to, back to our previous point, it's no different than uploading Mike Crispin's name to box because Right.

If the, if the terms of service are honest, they're not training on my name. 

Nate McBride: That's exactly right. So it doesn't matter. So the terms of service are honest. Exactly. Yes. And that's a premise that we have not talked about in 20 years. We, since SAS has come out, we've just assumed that terms of service are, are good.

Now we know that every single time there's been a breach of a major company that they did have our data 'cause someone got it. But we have no evidence of anybody having used our data to do other work. So now I, I take that back, there's a caveat to that. There are companies that will right size, for instance, if you buy a Google Workspace license and you're allocated a certain amount of, um.

Of terabytes to your account? Let's, let's say, yeah, that's actually not true. Let's say it's, uh, I [01:57:00] have Google I my own personal Google, Google license, and I have two terabytes of space. Google never actually gives me two terabytes. What they do is they gimme zero space and as soon as they add something to it, they, they add a little bit more space in front of it.

They're always working right in front of me. So maybe I paid for two terabytes, but I only have enough space for what I have today, plus a little bit more On the other side, there's always a buffer. They know this because they're training their model to always watch how much I have and then always increased by a small fraction.

That's the same for a lot of different, a lot of different situations with, with platforms. They know enough about you to make sure the environment's operating efficiently, but they don't know exactly what's in the data according to their terms of service. That's right. But even still, can you, can you imagine if the platforms allowed you to, and of course X grok does this, which is great.

Gr lets anybody of course, um, poison the well, [01:58:00] which leads to all the anarchy that's happening with grok. But imagine a model just, yeah, yeah, just go ahead. Whatever you type the prop we're gonna keep, it would be the end of that LLM in a day. Um, yeah, 

Mike Crispin: I mean that's some of the conversational memory stuff that you see in some of the, the tools, right?

Yeah. Is the memory component. 

Nate McBride: But still, even Claude, even Claude memory is only supposed to be short term if you read the for the session. That's right. But Claude now also has a recall, so you can actually refer to other conversations, which is actually, I think a better technique. Yeah. They seem to 

Mike Crispin: have their stuff together.

Nate McBride: All right. So do you wanna do some, um, predictability about where this all ends up or, 

Mike Crispin: sure. Yeah. I'd happy. 

Nate McBride: Well, yeah. Um, I wanna just crystal ball. This is 2028. So I was thinking about this today. 

So I was talking to company. It's always fun to do. 

I was talking to this company as I was consulting for them.

This was the second of [01:59:00] six classes. And, uh, as I'm, I'm making notes during the thing. I was thinking like, how different is my class gonna be in two years, like when I'm teaching it, if I'm still teaching it? Right. Um, what am I gonna be talking about? And it occurred to me that some of the things that are important to what we're talking about come from those notes for centralized.

Um, we'll be talking about vendors now. I'm not talking about vendors. Like who all the vendors are. The vendors. As in, you got a rep. I have a rep. We talk to our reps constantly. Our reps actually seem to give a shit. Um, there's more than two for the entire company. And we have relationships and they're not, and they're not bots.

They're not bots. They have conferences, they have user groups. They actually behave like vendors because name a single, a single vendor that operates like a normal SaaS vendor. None. They, they got, they got [02:00:00] nothing. They have no, no customer infrastructure whatsoever. None of them do. Um, they don't have time.

So, so in 2028, my hope is that I can slack my Claude person. Hey, what's going on? Uh oh. Just saying hi. Um, like I do with my poor ba the, the poor Bastar at Box, who's my rep who every day he is like, I'm like, Hey man, uh, Google's ripping off canvas. This is shit. He's like, oh my God, why does Nate keep talking to me?

Um, then okay, so, so the other thing was security controls. So we're, as of, um, Thursday of last week now using SSO under Okta for Claude. Mm-hmm. So, yay for us there. [02:01:00] Awesome. I, I don't see that changing so much, but I do eventually see DLP coming in at the, um, administrative level. Do you, do you think the same?

Yeah. 

Mike Crispin: Yeah. And I think the zero trust component of stuff is gonna become the shared key model. I think it's gonna get more popular right now. That's, you know, build it in Amazon, build it in Azure, or GCP is the only real way to, to get there. Or, uh, I think, you know, OpenAI has a shared, uh, IKE key model if you, if you use them, but that's it.

So I think those will become more, that'll become more prevalent across all the platforms. But I do also think that the coding agents, I like what. I think it might get more common for Codex and for other development tools to branch off into different license groups instead of an all you can eat model and the other vendors like Claude splits that up.

You gotta pay extra for that, which is a good thing. 'cause you don't want everyone in your company to [02:02:00] have that. Right? 

Yeah, 

that's a good thing. You know, if you, if you invest in CHATT Enterprise, you're getting Codex as part of the license and anyone who knows it gonna be dangerous. Kodak cameras. Yeah.

Kodak cameras. That's pretty cool. Uh, code. Yeah. Codex, uh, the, the, the coding. I know what you're talking about. I know what you mean. I know. Uh, but, um, that, those are the things when people start watching YouTube videos and figuring out how to use them, that could be Yeah. Pretty damaging to your, to your, to your business.

If they, they can get pretty dangerous to that stuff and so can a bad person, a bad hacker that's somehow has access to a machine 

Nate McBride: state for the record that if you ha, if you've looked at the c, the Anthropic or rather Claude Administrative console recently and the Gemini, sorry, the Google Workspace Gemini administrative console, yeah.

The options are very, very limited in what you can turn off and turn on. Yeah. I mean, that has got to change. 

Mike Crispin: Yep. 

Nate McBride: There is some shit that's [02:03:00] gotta be turned off and dis and give me the capability to disable, enable. They're not, they're not trying, or they're working on it somehow or they don't have any people working for them that can do it.

I don't know what the deal is, but the administrative controls are extraordinarily lacking for having been in the market this long. 

Mike Crispin: Is that you talking about Google or philanthropic? Both. Oh, both. Yeah. If you 

Nate McBride: go into Google Workspace and go to the, uh, gen ai, the Gemini, there's 

Mike Crispin: two 

Nate McBride: options. Yeah. It's like, turn this on.

Mike Crispin: That's one of the things they're extreme Gemini's extremely lacking on is the, is the ability to, to just create a project like just a shared space. I 

Nate McBride: Well, 

Mike Crispin: you have gems. Yeah, you do have gems. 

Nate McBride: But chat, that's not like apples to apples. It doesn't compare to a Claude project. But then Claude, Claude has artifacts, which are supposed to be like gems.

And so I don't really understand Claude artifacts necessarily. Like I've made a few little stupid ones and I'm like, [02:04:00] okay, um, now what do I do? Like, okay, good for me. Um, it's, uh, maybe somebody in my company will come up with something cool. 

Mike Crispin: And then you have someone in your company who decides to use opus all day and the whole fricking thing shuts down.

So 

Nate McBride: yeah, I always get the question like, so what's the difference between the extended thinking and research? I said, well, research takes a long time. It's gonna write a plan and it's gonna have you approve the plan and then do its thing, and you go back 15 minutes later and it's kind of almost done. And so on and forth.

I said, extended thinking just takes longer. It just, it does, it does the prompt, but it just takes longer to do it. It's still the same probabilistic. It's still doing probabilistic thinking, but it just says, it just says thinking on the screen. And I show them this in class. I'm like, okay, I'll turn this on, watch what happens.

And it just says, I'm thinking about it.

Mike Crispin: It's reasoning, Nate. 

Nate McBride: Its reasoning. Extended thinking. [02:05:00] Extended thinking. So why not just make extended thinking the default? It doesn't make it anyway, wants to wait. I love the fact that when you go into the, the, the Claude, um, prompt interface, you can still pick like sonnet three. I like, no, I'm not using the fucking new version.

I'm going back to the beginning. I want my prompts to suck. 

Mike Crispin: The, the other thing I can think of with that, and it's so I, I'm not sure it's the case in Claude, but in, uh, open ai, like if you want a a million, uh, token context window, you have to use GPT-4. You can't use five yet. 

Nate McBride: It's 200,000 no matter what.

No matter which one you use inside of Yep. 

Mike Crispin: In Claude 

Nate McBride: and I Have you ever, have you ever used Haiku, by the way? Yes. 

Okay. What's your experience been? Because I'll tell you mine, I, 

Mike Crispin: I, I like, um, just I, I'll type very basic stuff in there because I need a quick response. Like, you know, uh, how many, how many, how many fingers [02:06:00] on a human hand, like that type of stuff.

Nate McBride: Yeah. That's, that's pretty much how I explain it to my classes because I'm like, I'm like, like, what's 

haiku about? I'm like, well, let's suppose that you just really wanted one answer to one thing in the world. You could use haiku. Well, can I use Saana for that? Yes, you could. 

So there you go. Uh, just don't, 

Mike Crispin: just don't let it, you, you could ask Haiku, uh, a hundred questions and you could, you can ask sonnet, you know, 30.

So it's thing, what I think they should do 

Nate McBride: is they should make honestly a haiku, it's a namesake. When you ask a question, your output is always in the haiku end. The story. Maybe 

Mike Crispin: that's what the original idea was, was, 

Nate McBride: yeah. They're like, well, we can, we can't do tweets, so let's see, what else could we do? Oh my gosh, that's funny.

Any other two? 2020. So first of all, in 2028, are we still gonna be talking about decentralized and centralized it, you [02:07:00] think? 

Mike Crispin: I think it'll still be the same as it is today. I don't think it's gonna change all that much. And I think it'll, it's, it'll be a, it depends, um, just sort, sort of question. I mean, maybe, maybe in 2038 it must be a much different story, but I think even three years from now, it's, it's, um, it'll still, still probably be very similar to the way it is now.

Nate McBride: What, what flavor, what flavor pro protein shakes are you gonna want drink in 2038? Like when you're sitting in your, your chair Yeah. Watching TV for 20 hours a day. What, what flavor of protein shakes are you gonna drink? That's a good question. You gonna vary 

Mike Crispin: it? Peppermint bark. So let me tell you, I got, I, I'm getting all these and someone must know me.

Well, I'm getting all these emails and text messages from Shake Shack. Telling [02:08:00] me that the peppermint bark, they miss you. Shake is available. Yeah. So, so me and Berdo go over to Shake Shack for lunch on Thursday? Or was it, what was it? No, Tuesday. Uh, and we're sitting there like, should we get, there's one of these fucking shakes, like, 

Nate McBride: yeah.

Yeah. They look 

Mike Crispin: amazing. Guess how many calories is in the, uh, the Peppermint Bark Shake Shack Shake. Now this is like an eight ounce shake. 1200, 1,380 1,380. That's your whole day. You're done. You don't have to eat anymore. Yeah. It's like, and that, the thing is, is I would drink that in about eight seconds.

Gone 1300 calories in, in less than five seconds. Est well done that. We didn't get one. We had self-control. We looked at them instead. We got like, I don't know, like triple burgers and mos of [02:09:00] french fries. Oh. 

Nate McBride: Because yeah, 

I'm sure that was 

like only a hundred calories. 

Mike Crispin: Every little bit helps, you know? Yeah.

And then, and, and then to, to kind of soften the blow. I got some unsweetened iced tea. 

Nate McBride: So I think, I think honestly what you should do, see, see if you could get 5,000 calories in one meal at Shake Shack. 

Mike Crispin: Oh, it's easy. Oh, we can do that easily. 

Nate McBride: Well, well, I, I, I wanna have the 5,000 calorie challenge. I'll do it with you, by the way.

Yeah. And we'll just, you know, order like a normal meal and see if we can get 5,000 calories. 

Mike Crispin: Oh, that'd be great. Now we, now we can roll around the mall afterwards. Go to Burlington Mall and just go to the Tempur-Pedic place and fart on all the mattresses. I think that would be awesome. 

Nate McBride: Vomit in the fake plants.

The Shake Shack, because now 

like the, the worst 

meat sweats all over the furniture. 

Mike Crispin: Oh, dude, you left sweat marks in the seat. It's like, oh. And then we can go to Annie's pretzels. [02:10:00] Yes. And get some big like sugar and cinnamon dipped 

Nate McBride: pretzels. Yeah. Guy comes over, he is like, uh, what can I get you? I'm like, I'll have the diet pretzel.

Mike Crispin: Yeah. 

Nate McBride: I just ate at Shake Shack. He's like, I get you. Let me get you one diet pretzel. It's only 90 calories. We couldn't, 

Mike Crispin: we could in one of the podcasts have like Arun a prune juice drinking contest until what happens. Just, we'll just keep it going. We have a marathon podcast. We drink as much p prune juice as we can, and whoever has to get up first, that's, that's who wins, who loses.

Okay. And right. And then, uh, let's do it. And then that'll be, that'll be the, and we can do it live on Twitch. We'll put it on, uh, we'll have like, like maybe one listener for about 20 minutes. Uh, I think we should, 

Nate McBride: well, it'd be great. We just do it from the toilet. We don't have to get up.

Mike Crispin: That would be the end of that, that would be the end of the podcast, I think if we weren't [02:11:00] that far. But hey, who knows? Maybe it'd become a TikTok meme or something. Who knows? 

Nate McBride: Could be two idiots drinking produce juice. 

Mike Crispin: You know, TikTok live. I have enough, I've, I've been on it enough that I have a live, I can do live videos now.

So why I wonder if our, what do you, I wonder if our, oh, I don't doom scrolling now I'm watching all sorts of crazy stuff on there. Oh man. Yeah, mostly about prune juice. My favorite prune juice is on there. They, they advertise it for me. Fantastic. I thought you working on like your 

Nate McBride: next, next rap album or something?

Mike Crispin: Well, that and oh, actually, uh, I'm also trying to get back into Fig Newton's. So the more that, the more fig Newtons I can eat, along with the prune juice that the news news 

Nate McBride: are, are a magical snack because they can save your ass from almost any disaster. It's like an MRE. So those stick with you, right. 

Mike Crispin: So 

Nate McBride: yeah.

Mike Crispin: Yeah, they stick to your ribs. They're, [02:12:00] yeah, they're dense. I love fake Newton's. They do stuff Underrated, underrated snack. I love fake Newton's too. I eat them during all my big long races. I bet. 

So that snack works as a fuel fueler for you. You just, you crush those. Absolutely. 

Nate McBride: Absolutely. Hey, so anyway, that was a good conversation about decentralized, centralized.

It, it's good to know that the more things change, the more they mostly stay the same, uh, at least in the terms of it, departmental models and structures. So, yay. Just an opinion. Yay for it. Maybe. Um, I bet you Gardner, we're still, I bet you Gardner wrote something or, or McKinsey wrote something that like, like what, like BCG wrote an article the other day about the a all AI company and else, you gotta be fucking kidding me.

Like, who, who, who's coming up with this stuff. But I think that, uh, I think that we'll find some, some model, probably the hybrid [02:13:00] thing somewhere down the road. And, uh, yeah, I mean, we're gonna, we're gonna come back to this idea by the way, of the, the function when we get back to eliminating the, uh, the eliminating It episode, um, which I think is episode.

Let's see, let's see, let's see. Did we, did we put that on the season? Uh, oh, well, episode 31 is why we still do things the old way, kind of a, um, a look back at our tropes from season one, but actually serious, uh, no, I don't, we actually don't have an episode on eliminating it. I'm just gonna have to weave that narrative in secretly behind the scenes when you're not paying attention.

Um, who's switching the backup tapes tonight?

Oh, man. Remember, we should, I think we asked Bob that when he was on, but that [02:14:00] was Bob's job at TKT doing the, the, the datto, the dad tapes and having like, just so many of them. Um, 

Mike Crispin: yeah, dude, I, I haven't done that in a month. I forgot. Sorry. There's, there's 600 errors in the net backup log. I forgot to put the tapes in.

Nate McBride: Delete that file. It will go away. 

Mike Crispin: Or we just kept putting the same tape in over, over again. Now that's okay. Like with Ignite Tapes or like if you're at HPUX or, I know just like the nightly thing, but it's not gonna, I think, I think, I think was, 

Nate McBride: Matt was a little high on the stress meter during a lot of those conversations we're just all like, no man, it's cool.

We'd you put the same tape back in. And he's like. Lucina shit. 

Mike Crispin: Oh man. What did you just say, man? I'm like, oh, I, let's just, you just take the tape out and put it back in. Right?[02:15:00] 

He's like, no, no. There's a box of tapes with a number on them. You gotta Yeah. Days of the week. Days of the week, man. Put the right one in. In my whole, 

Nate McBride: my whole career, I've never seen a data center that looked as good as his. Yep. Nor better organization of the things in it. Cables. Yeah. I mean, it was art.

Mike Crispin: Yes. 

Nate McBride: This should be. You spent a lot of time 

Mike Crispin: wrapping those cables. 

Nate McBride: Yeah. It was an art. They were all also perfect. Everything measured perfectly. Same distance. Yep. So next week, verification economy, baby. Let's go. We're here. I'm stoked. Let's go. Let's go. It's gonna be a good one. 

Mike Crispin: Um, there's a lot of different like definitions and ideas of what that actually means online.

Like if you look it up, it, it's all over the map. On, on, on the line. Yeah. On the line. Go on the line. Check that out.

Nate McBride: Well, [02:16:00] I mean, you have some wonderful takes on this. Uh, I don't know if I have any wonderful takes on it, but I definitely have opinions on it. Um, but we're now, I mean, between. Did you read the article about Reuters the other day? Thomson Reuters, about how much data they own, about people and how they've been selling it.

Mike Crispin: Was that, was that on 4 0 4? I think I saw that. It 

Nate McBride: was, it was, 

Mike Crispin: yeah. 

Nate McBride: I did see that. Yep. I didn't know this. That was shocking to me to read. Uh, where was that? Um, 

Mike Crispin: it's, it's a lot, a lot of it is related to the, the World Check database they have for vendors. Yes. Like, I think it's all related to that because of the, you know, to make sure that there's no, uh, you know, criminal connections or any sort of connections to anything that would put your company at risk doing business with certain ve certain companies or, you know, [02:17:00] relationships to laundering organizations or past criminal history or whatever it might be.

But I believe they have co they have part of that data collection. Can, can, has arms that reach out everywhere, so. 

Nate McBride: Yeah. Yeah. And they've been selling, selling it. It's pretty interesting. Which is the crazy part. 

Mike Crispin: That's a world, it's World Check I think is what it's called. Yeah. Because I've used that before.

Nate McBride: So 

Mike Crispin: I mean, it's part of another company. 

Nate McBride: Yeah. Real ID and clear, clear Id, all that shit. My God. Um, anywho, we can talk about all that stuff too, because, well, it's identity and, um, it's going away. Um, yeah. All right. Anything else before we break camp? 

Mike Crispin: No, no. Good episode. Lot of good discussion, and I think it'll be good to talk next week.

And near and dear my life. I think this is like one 

Nate McBride: of those, one of those topics that people should be thinking actively about when you look at your [02:18:00] department and where you're going and how you're going to evolve it and how you're going to respond to where AI fits and all that stuff. This is, uh, this is right at the top of the pile in my opinion.

Um, agree. All right. Well, uh, to next week we will be taking a break because we are gonna go eat mashed potatoes by the pound. I'll be back. Yes, I love mashed 

Mike Crispin: potatoes. 

Nate McBride: We will be back on December 3rd for our, um, post Thanksgiving Day, pre-Christmas, um, episode. Excellent. Can't wait. We'll be talking about verification of the comedy at that point in time.

Uh, so it's be especially nice to, um, old people and, um, pets and young people and people that are in between. Make sure that during your Christmas or Thanksgiving dinner you talk about politics. That way you can reduce the number of presents you have to send out this Christmas. Um, make sure that you're extra thankful for the podcast at your Thanksgiving celebration.[02:19:00] 

Um, however you may enjoy your Thanksgiving celebration or if you completely, um, are opposed to the idea of Thanksgiving, then whatever you do on that particular day, just. Think about us for a moment. Um, have your pet spa or neutered. Uh, don't be a dick to it. Uh oh. We have a website. The C oit Us. Yeah. Uh, all of our, we do have a merchandise shop.

It's in our, um, podcast links at the bottom, as well as our substack. We have a slack board. It's kick ass. You can get an invite to it from the bottom of every single one of our posts. I feel like I'm missing one. Oh, and give us all the stars. Thanks. Yeah. All the stars. All the stars. All right, Mike. Good to see you.

Rock roll man. This was way better than going to see wicked. 

Mike Crispin: Yes. 

Nate McBride: Part. I, 

Mike Crispin: I agree. I agree. [02:20:00] 

Nate McBride: All right. This is great. You're cuter anyway, and I like, I like looking at your basement ceiling. 

Mike Crispin: Yes. This is, this is true character. Can, can you imagine? Can you imagine being where I am right now? Amazing. Hollywood.

Hollywood 

Nate McBride: can't do this. I mean, this isn't AI generated. Look at, 

Mike Crispin: even the camera knows to go up there. Look at that. Yeah. It automatically, yeah. This is just a nice, I like, I just want you to know that this is all intentional. It's a wonderful design that I've created here. Oh my God. It's very, 

Nate McBride: to remind you, by the way, five months to bio it world.

Just mark your calendar. 

Mike Crispin: Oh. What are we gonna get on some panel? I wanna get on a panel again. Can we do that? Yeah. We'll be on a panel. Oh, I gotta find, we gotta find a good one. We gotta find a good one. Digital. Let's try and get on two panels this time. Two panels. All 

right. Alright, dude, I'll talk to later, 

man.

Be good. Take care, Nate.[02:21:00] 

Trance Bot: The calculus of it,

season three,

verifying this identity.

Sometimes you just 

have to take it.

Sometimes you just have to take it

because it's season three divided autonomy,

verifying identity,

the calculus of [02:22:00] it.