[00:00:00] Coté: Now, Whitney, have you ever encountered an enterprise architect

[00:00:04] Whitney Lee: Yes

[00:00:06] Coté: And what, what, what was, what's, what's your impression of, of what an enterprise architect does in, uh, what is this? 2026

[00:00:13] Whitney Lee: Uh, well, first of all, I don't have 2026 experience with any enterprise architects.

[00:00:17] Coté: But in the past

[00:00:18] Whitney Lee: stale. But yeah, um, my impression of the enterprise architects is, um, they seem overworked. They're too busy.

[00:00:28] Coté: Mm-hmm.

[00:00:29] Whitney Lee: a sales meeting, they, um, listen to a short thing and s- they say the minimum they need to, to move the ball forward on that particular thing, and then they gotta hop off to get to their next thing

[00:00:39] Coté: Oh. Oh, I love that phrase. Is, is that, is that like an American meeting phrase, "I've gotta hop to another call"?

[00:00:44] Whitney Lee: off,

[00:00:44] Coté: Hop off. Yeah, hop here, hop there, hopping everywhere. It s- it seems like, you know, I, I remember, uh, I've encountered many enterprise architects in, in my career, especially back in the days of, of what they used to call Java Enterprise, which was all about enterprise architects, I think.

[00:01:03] Java 2EE? I, I forget what it's called. But, uh, uh, I, I think a little while ago, uh, I even had a presentation about enterprise architects, and it was kind of hard to figure out how they, they fit into things or what their role was. And so, I mean, your encounter with them sounds like they're busy doing something, but I'm not, I'm not quite sure.

[00:01:27] I don't know how the role has changed from, like, planning things out. Like, one time I met an enterprise architect, and it was one of those, like, the, you know, we, we, we all, uh, work at, at vendors. It was one of those meetings you have where the person really just wants to let you know they are not interested in you, but they are also very smart.

[00:01:48] And, and so they, they, they, they want you to just be there, and they wanna talk with you and tell you things and go over stuff, but they have no, no reason to want anything that, that you might be representing for them. But there was-- They did have this great almost wall-sized whiteboard that had their entire architecture for their company on it with all the different boxes and systems, and that's kind of what, what I think of an enterprise architect is, is like, is like the person who draws that.

[00:02:20] Uh, and, and, you know, that's, that's like an artifact creation way of doing it. Like, what the, um... What's another American businessism? The, the outputs? The outcomes? Deliverable. That's, that's what they say. That's, that's their deliverable. But it feels like

[00:02:37] Whitney Lee: I didn't

[00:02:37] Coté: there's something more

[00:02:39] Whitney Lee: were uniquely American.

[00:02:40] Coté: Deliverable? I don't know.

[00:02:42] I f-

[00:02:42] Whitney Lee: popping in and off of, uh, in and out of a meeting

[00:02:46] Coté: Yeah. Yeah. I feel like, you know, Francis Bacon and Pascal didn't say deliverable. Like, it's probably got created over here. I'm, I'm, I'm, I'm not sure. They, they probably didn't... And, you know, Luther probably wasn't like, "Hey, Pope, here's my deliverable. Uh, and, and I've gotta go pop off somewhere else and do something."

[00:03:07] Which I gu- I guess it's either, it's either 17th century philosophers and religious people or American. That's pretty much the, the divide there.

[00:03:17] Whitney Lee: ping you about this.

[00:03:20] Coté: Uh, we, we'll do a double click on it. Uh, well, well, let me,

[00:03:27] Sorry.

[00:03:28] let, let me, let me... Speaking of zooming in, let me, let me ask our guest, Rick Clark. We used to work together.

[00:03:34] As we were just remembering, we used to, uh, enjoy the same, uh, fresh hummus and great, uh, coffee here in Amsterdam. What is, what is... It's, it's 2026, September 1st as we record this. What is an enterprise architect now?

[00:03:50] Rick Clark: Um, well, it is a, an exceptionally rare thing that you don't see very often that is nearly extinct. And the things that we call them now are not what they used to be or what they should be. So w- when Whitney was describing the enterprise architect, it sounded more like a solution architect popping into a meeting, 

[00:04:10] Coté: hmm. 

[00:04:10] Rick Clark: leaving. Um, and Coté, when you described the enterprise architect, you sounded like the old enterprise architect

[00:04:18] Coté: Oh, right, right

[00:04:18] Rick Clark: was like the... There's a reason we don't have the, the The enterprise architecture that we used to have, and that's the reason. They pop into a meeting to show they're superior to you, and

[00:04:32] Coté: Mm-hmm.

[00:04:34] Rick Clark: uh, and we-- So we lost enterprise architecture as we shifted all the decisions left, right?

[00:04:40] Coté: Mm-hmm.

[00:04:41] Rick Clark: part of what, what's happened over the last 15 years is we shifted left, we shifted left, we shifted left, and we, we thought, we decided, ah, um, developer flexibility equals developer productivity equals value, right?

[00:04:55] Mm-hmm.

[00:04:55] these, uh, uh, architectural board review, this architecture, you have to do this, you have to do that, that slows down developers. They have to ask for approval for everything. So we carved out en-enterprise architecture. It slowly disappeared. It became first marchitecture, and then it disappeared in, in a, in a lot of places.

[00:05:14] It, it disappeared, and, you know, it left us with what I'm going to describe as chaos. No architectural coherence, right? Whi-which, you know, we, we, which we celebrated. Woo-hoo. Developers now can do everything. Now developers have to know everything. We're full stack developers. We deploy our own infrastructure.

[00:05:37] We build our own development pipeline. We make our own rules, and nothing slows us down. That's a lot of cognitive load, though. is a lot of cognitive load that we have shifted to the developer, and it's also meant that there's no decisions being made. There's

[00:05:52] Whitney Lee: What?

[00:05:53] Rick Clark: architecture

[00:05:54] Whitney Lee: hear you, but isn't that exactly the problem that like platform engineering is meant to solve?

[00:06:00] Rick Clark: It is.

[00:06:01] Whitney Lee: Uh-huh.

[00:06:02] Rick Clark: it had only done that. Um,

[00:06:04] Whitney Lee: Uh-uh.

[00:06:04] Rick Clark: so yeah, so platform engineering ends up being self-service flexibility. That's what we-- That's what, that, that's what it's become. Now, you go back to Cloud Foundry, and it was trying... It, it was not that, right? It was trying to be, um, an opinionated platform that you, had intent, and then the platform could or deploy that intent. But, uh, that's not where we ended up with things like Kubernetes. Now we declare the how, not the what. Build this, build that. Now I, developers are, are, are still building Helm charts and, uh,

[00:06:45] Whitney Lee: So the, the problem...

[00:06:47] Rick Clark: those sorts of things

[00:06:48] Whitney Lee: What I'm hearing you say is the problem is that they're forced to have an opinion about these things that are available to them self-service. Like ideally, they wouldn't have a opinion about which relational database they use. Like they shouldn't need to, to worry about that. They should focus on what they do that adds business value

[00:07:06] Rick Clark: E-e-every time a developer is, is worrying about Terraform or worrying about infrastructure or worrying about what kind of database, that's not, that, that... They're not writing logic that generates value. Like that, that is... So y-yeah, I, I think they shouldn't be doing that. And I think that that's cognitive load

[00:07:28] Coté: Mm-hmm.

[00:07:29] Rick Clark: that is difficult for, for any, any human to handle. Um, and I, I think that it-it's become sacrosanct. It's become religion. We're not really allowed to question that. Developer flexibility is developer productivity, I don't think it is. And, and I think that, you know, AI forces us to rethink that

[00:07:54] Whitney Lee: I have so many questions. One of them is, uh, this marketecture term, but really what I, I wonder, wonder, okay, devil's advocate, is there... Are you making a false dichotomy between a, a developer having zero choice or a developer choosing every single thing? Like, are there maybe a few things, options they should have? Like, what if they just have a few? Is there any benefit to that? Are you just saying no?

[00:08:23] Rick Clark: Um, there's a benefit to that if it helps them accomplish their goal and it

[00:08:26] Coté: Uh-huh

[00:08:27] Rick Clark: other goals. So yes, there, there should be. Like how they write their code, th- there, there should be. There should be some choice. But that choice should be what normal developer choice, right? They're

[00:08:42] Coté: Mm-hmm

[00:08:42] Rick Clark: g- optimizing things, they're worried about scalability, they're worried about, you know, things, things in their code. It should be about business logic. It shouldn't be... So here, here's, here's a good example, right? PII, right?

[00:08:55] Coté: Mm-hmm.

[00:08:55] Rick Clark: I... Do I need to define PII? That, that's, that is, that is personal information, right?

[00:09:01] Mm-hmm

[00:09:02] that is generally regulated, and is regulated differently everywhere, right? As a developer, I shouldn't need to know what rules need to be in the database.

[00:09:12] I shouldn't need to know how I need to build a database for regulatory compliance of PII. should say to the platform, "I have PII,"

[00:09:22] Coté: Mm-hmm.

[00:09:23] Rick Clark: it should just happen, right? And I don't... W- why would I care? How am I doing a better job of writing my by building the database myself? Like all, all I've done, all they've done is m- made me do more work at, at the end of the day, right?

[00:09:42] Whitney Lee: I feel this way every time I'm asked to sit through an OKR meeting where my OKR took 30 seconds of that hour-long block. I, I think that I-- Mm, like maybe it's good to understand where the team's going, but, but since it's not I can contribute, then it's really taking up cognitive load and literal time. Um, so I wanna hear what is marchitecture real quick before we move on.

[00:10:10] Rick Clark: first of all, I- first person I heard use that was Ed McLaughlin, the President of Technology at MasterCard. and it is basically where, uh, you- you're, you're, you're not actually enforcing any architecture, you are creating pretty slides that describe things, and that's the- that's your role. you're showing, "Look at how beautiful our architecture is," and he called that markitecture.

[00:10:33] And I kinda like the word, so I, I stole it.

[00:10:36] It's sort, it's sort-- You're, you're like the mapmaker, the, the, the map of the system, the cartographer of, of how things are. And, and, and, and I guess-- And th-this, this gets into the other thing of the... It, it, it seems like part of, of that marketecture, enterprise architecture is not only are you the mapmaker, but you're trying to feel out how you can do some terraforming with your pencil.

[00:11:01] Coté: Just sort of like the, the way that we draw this map, I would like it to actually be this way. And, and, uh, you know, that gets us closer towards, I think, like, like, what, like, what should an enterprise architect be? Like, we talked a little bit about platform engineers or the platform, but I feel like an enterprise architect is somewhere else in, in this situation

[00:11:23] Rick Clark: So, so we have to make a distinction what inter- enterprise architecture was and what it should be this, in, in the modern world, right? What it was, was not really-- There wasn't platform engineering back in the day. And i- if you're not, if you're listening to this, you can't tell that I'm, I'm old enough that I remember those days, um, when there were enterprise architecture teams.

[00:11:45] And what it was back then was, uh, a bunch of people that used to be developers but now thought they were better than you, and that you had to go present your designs to so that they could show how better they were than you, and you c- and tell you to make changes so you could get approval. And that did s- it slowed things down, and it was arrogant.

[00:12:04] And, and I, I understand why we pushed past that, right? I understand that having a gate of people that, that are, you know, the lords of technology that, that you have to go show your work to, that, that, that, that didn't work. But what they should be doing is they should be defining, uh, what architectural coherence means.

[00:12:26] And I,

[00:12:27] Coté: Yeah

[00:12:27] Rick Clark: you, I'll give you an example. I'll give you an actual real-world example, okay? I, I talked to a, a large insurance company, and, um, as part of this workshop we were doing, we discovered they had over 50 secrets managers in production, right? So how many, how many, like, valid use cases are there really for a secret manager? I, I could come up with a few, and that's m- that's, that, that's pushing it. Um- But the reason they had 50 secrets managers is because everything was pushed to the edge, developer decisions. And the developer decision often was what was easiest to do at 2:00 AM before the deadline,

[00:13:09] Coté: Mm-hmm.

[00:13:10] Rick Clark: And that, that is, that is-- At that point, they had wanted to do governance as a service. Well, that was hard. Trying to do governance as a service when you have absolute chaos and anarchy is, is difficult. Now imagine you, you put AI into that. What is, what is correct? What does AI infer from, right? There-- th- what, what architecture should do is they should be supplying the rules for the platform. you say, "I have PII I need to store," right?

[00:13:42] Coté: Mm-hmm.

[00:13:42] Rick Clark: Then they should-- the, the platform should automatically know what to, what to do with that. If you have something that is very specific, "I have PII, but it's in, uh, China or

[00:13:56] Mm-hmm.

[00:13:57] India or it's, it..." Then-- And that use case isn't covered, the enterprise architects should figure out how to cover that, and that should then be in the platform.

[00:14:05] I imagine in the future this is all just they're controlling the rules that agents apply, and if agents can't apply the rule, it kicks back out, or the, the ca- use case doesn't fit the rules, it kicks back out to the enterprise architect who says, "Either we're gonna create a new use case and new rules for this, or no, you're gonna do something else."

[00:14:25] So I imagine it being that they control architectural coherence in an enterprise. And they have to because AI isn't g- is going to do whatever it wants. Like, someone has to be controlling that, that, that architectural coherence.

[00:14:41] Whitney Lee: So how do you get from, pre-AI already it's a mess because there are 50 different secret managers. Every developer team is making their own. They're unaware of what the other people are doing, and that is compounded by all different technologies that they're each choosing or not choosing, um, and only operating within themselves.

[00:15:00] And now we have AI. do you get... First of all, what's AI doing without the proper guidance when, when you don't define that? And second of all, how do you get to a place like what the utopia you described where

[00:15:14] Rick Clark: Yeah.

[00:15:14] Coté: Coke or yes

[00:15:16] Rick Clark: what AI is doing is it's failing. There, there's a reason why 88% of AI projects and POCs don't actually go to production,

[00:15:25] Coté: Mm-hmm.

[00:15:25] Rick Clark: work in their little POC environment, and when you release them into the chaos, they don't. I mean, the, uh I

[00:15:34] Coté: Hmm.

[00:15:35] Rick Clark: of, of, of this is it's gonna be expensive. It's gonna be expensive to fix this thing. You, you, you spent 15 years making mistakes, and you're

[00:15:43] Uh-huh

[00:15:43] f- you're gonna have to bite the bullet and you're gonna have to fix it. You, y- AI's not gonna fix it for you. I mean, the-- certainly AI could be used to discover them all and help you come up and, and, and figure out, hey, we need these in these three model...

[00:15:57] AI could be a part of the solution, but you're gonna have to pay to fix it. You're gonna have to fix this mistake you made. You dug yourself into a hole, you know, there's-- you have to get out of the hole. And, and that, that is I think where a lot of, of companies are gonna be they're trying to-- when they are failing with AI.

[00:16:17] They're gonna find out that, ah, we need to start... We need some coherence first

[00:16:22] Coté: Well, well, let, let's, let's talk about this whole escape velocity, 'cause this, this is always a, a, a, a favorite pondering is... We'll make an analogy. Like, I kind of know I should probably check, take my car in to get it checked out, right? Like, it hasn't been in a while. The, like, that tire signal is on. I know it's broken because it's been on for, like, a year and nothing has happened, right?

[00:16:45] So, like, obviously it's fine. But, like, I'm not really gonna take my car to get it checked out until it breaks or I have a reason, and then I'm gonna be like, "What the, what the fuck was I doing? Why didn't I just take this car in?" And, you know, the reason is 'cause it's a lot of effort, blah, blah, blah, right?

[00:17:02] And, and, uh, and you know, a lot of, a lot of the being on a whole feels kind of like that in the, the, uh, you know, most people, most people, I don't know why I'm saying it like that, they, like, know that we should not be in this shitty state and, and we should do the work to do it. And unless, it feels like unless there's some crisis or, like, um, let's say reason to do something, they'll just not take their car to the dealership.

[00:17:31] And so, like, it does seem like... I, I, I, I don't know. Like, it's, it's this really frustrating situation where, like, both in the car situation, which I don't know, maybe it's just quality of life. I don't know if that's frustrating. But in a work situation, it's like everyone knows things are screwed up, but then the fact that they don't do anything, maybe they're not screwed up.

[00:17:51] Like, what-- how did they get into this state?

[00:17:53] Rick Clark: You, you have

[00:17:54] Coté: I

[00:17:54] Rick Clark: idea how apropos that, that you-- So the next thing I'm doing after this podcast is taking my car in. And, and, and, there's a reason for it. So I, I scraped it on a, a, a pole in a parking garage at Whole Foods, and, um, what I needed was a catalyst. It's very-- It's a tiny little scrape, but it's a leased car, and in three months I have to turn it in. And that, that

[00:18:19] Coté: Ah.

[00:18:20] Rick Clark: and that, that cost I'm gonna have to pay if they look at the scrape is the catalyst for me to go get it fixed.

[00:18:27] Coté: Mm-hmm.

[00:18:27] Rick Clark: AI is the catalyst for, and the money spent on AI is the catalyst for fixing your, your architecture. I mean, peop- y- you, you're-- Everyone is dumping tons of money in AI. A lot of it

[00:18:41] Coté: Yeah.

[00:18:41] Rick Clark: fear, right? They're doing it out

[00:18:43] Mm-hmm.

[00:18:44] Fear that another company is going to do it first, and they're gonna have to be dragged into their board of directors and say, "Why did we lose market share? Why aren't you guys ready for this?" So it's, it's driven by fear, but it's a massive amount of, of investment, right?

[00:18:59] So you're gonna be spending your hundreds of millions, billions on AI, spend, spend a few million fixing your architecture. That's, that's the catalyst. And

[00:19:08] Coté: There's-- It's

[00:19:09] Rick Clark: that, you're gonna be, you know, gonna be one of the failures. And we're, we're entering an age where you're-- think we're going to see of displacement.

[00:19:22] Coté: Oh yeah

[00:19:22] Rick Clark: don't do this right, you're going to be displaced. that, that's what I think is, is coming

[00:19:30] Whitney Lee: I think, I think fixing stuff is not very sexy. Yeah, like pre-AI even, like, um, in politics, you don't see a politician running on, like, "We're gonna fix the few cracks there are in the highway." Like, that's not, "I'm gonna build a new highway," is amazing. "I've left a legacy." Like, "I've, uh, lightly repaved what was already there and did some needed maintenance," nobody was like, "Woo." And so I think the sim- same thing in business. Like, I've, I what we already have a little more secure, a little more, like, resilient. I m- I... Like, we improved it, we solidified it, uh, like we, um, cohered it.

[00:20:11] Rick Clark: know, so, uh, that, that's an- another good... Calling it sexy is another, another good analogy. That,

[00:20:15] Yeah

[00:20:16] So maybe you need to brush your teeth if you wanna be sexy. Maybe you need to take a shower, get the stink off you. So what, what I-- The way I like to, to do it is if I... So I, I do advisory work, right?

[00:20:29] I advise people, and some of the

[00:20:30] Coté: Uh-huh

[00:20:30] Rick Clark: people about is how to get to

[00:20:32] Coté: Yeah

[00:20:33] Rick Clark: I don't go in with flashy AI things. I go in and say, "Well, this is what you have to do first

[00:20:39] Coté: Mm-hmm.

[00:20:40] Rick Clark: succeed," right? If you wanna succeed, you need observability because you don't need... You, you need real observability, and you need architectural coherence, they're a small part of what you're going to have to build, but if you don't build them, you will fail.

[00:20:57] Coté: Mm-hmm.

[00:20:58] Rick Clark: you bring in the, like the, the MIT study from last year that says 95% of, uh, AI investments don't return anything, right? Um, there's, there, the, tho- those are some of the reasons. So yeah, you have to... You're gonna... You have to brush your teeth. You have to

[00:21:14] Mm-hmm.

[00:21:15] You have to put on some deodorant.

[00:21:17] You, you, you, uh, you have to not be stinky and, and that, that, that makes you a little more attractive. Um, z-

[00:21:25] Whitney Lee: And to

[00:21:26] Rick Clark: and

[00:21:26] in particular, that's who I'm trying to woo

[00:21:28] you, you have the, the, they want AI and they don't wanna fail, and they're afraid to fail, and they're afraid to, to fall behind others. And, and if you're, you're someone like me that's advising them, you have

[00:21:40] Mm-hmm.

[00:21:40] about that.

[00:21:41] You have to be, "L- you're not ready." you know, "Let's do this POC, but at the same time you need to start fixing these things or you're gonna fail."

[00:21:51] Whitney Lee: Are, do you find people are hearing you when they say that, or are they too focused on the shiny thing?

[00:21:57] Rick Clark: Oh, no, they're, they're mostly not hearing me.

[00:21:59] Whitney Lee: Okay.

[00:21:59] Rick Clark: yeah, but I'm used to that. I'm, I'm used to tilting at windmills and, and no one giving a damn.

[00:22:04] Coté: Well, th- that's, that, that's a fun little, little side topic is how do you, uh, you know, to use the windmill thing, Don Quixote was insane, so, like, that's how he dealt with this. Like, he was living in a fantasy world, which is, s- sometimes seems like a great option, uh, you know, if, if you're in his situation.

[00:22:23] But, you know, if, if you're in it and, and knowing, knowing your history as well, like, if you're in an advisory role and you feel like, uh, people aren't listening, doesn't that, like, drive you a little crazy? Like, like, and what do you, what do you do about it so you don't, like, kinda turn into, like, either, like, really depressed or, like, crotchety?

[00:22:43] Like, there's gotta be some mental fortitude you have to cohere up there in- into happiness

[00:22:50] Rick Clark: That's interesting. I, I, I don't know. I, you know, perhaps I am in- insane. Um, we, we did a lot of, uh... So Coté and I did a lot of, um, the CxO dinners and breakfasts together years ago, right? And, and y- as we did those, what I would do is I would start telling stories, I would wait until someone acted like the story m- hit, hit, hit something inside them and responded and started talking back, and then we'd start talking about that thing. So I generally try to find a pain point or something that they can associate with, something they can relate to and, and, and, and use that. And use that to hopefully make them listen to me. But I just think, I guess I'm, uh, probably a little crazy. I just think I'm right. um, and, and, uh, like, and the, the salespeople don't like you because they want to sell things, and you're in there

[00:23:47] Coté: Mm-hmm.

[00:23:47] Rick Clark: telling the customer something they don't wanna hear,

[00:23:50] Coté: Mm-hmm.

[00:23:51] Rick Clark: I honestly think that it's my job to tell them the truth, and sometimes when you tell someone the truth, they don't believe you right away. It takes a few years of them doing the wrong thing and then they'll remember, guy told me the truth."

[00:24:07] Coté: Mm-hmm.

[00:24:08] Rick Clark: that's, that's how I look at it. I feel like one day they're gonna know I was right. Um, that, that's,

[00:24:13] Coté: It's, it's, it's a, it's a long game is what you're saying

[00:24:16] Rick Clark: I... Well, we had a, we had a similar situation, uh, and I... You might have been there for this. We were at some large, I believe, automobile manufacturer in Germany, and, uh, he, he told us, um, he ha- his developers have to have every, they have to have every single native, uh, feature on every single cloud available to them at all times or it will slow them down. And I think the result was, "Good luck with that." And, uh, and then we laughed

[00:24:48] Coté: Yeah. You know, I, I, it's, it's, it's-- There must be a similar thing. Like, I, I remember one time I was, I was, uh, uh, talking to my therapist, and I was like, "This person must be so bored," right? Like, they h- they hear these same problems over and over again, especially from me. And, uh, you know, like, I'm, I'm sure they have the same conversation.

[00:25:07] And there must be something... I mean, 'cause it's not like a therapist gets paid a lot of money, right? So there must be some liking of, of, of that kind of situation that keeps people in that role. I'm, I'm, I'm not sure. And similarly, if you're in an advisory role, a lot of what you're doing is not getting hired.

[00:25:31] And, and, and maybe also, like, like, like, like an e- even, even bad situation, like wh- on, on the other podcast, Brandon tells this story all the time. He used to work on modernization, and you go, and you go and talk with people in large enterprises, and they're like, "Oh, we gotta modernize whatever. We gotta do our modernizing."

[00:25:48] And then, you know, uh, nothing ever happens, right? Like, they talk about it, they want it, but then it just other things come up, and so it's not a priority. And it seems like you have to build up the ability to just, like, deal with that. And, and, like, kind of either you don't care about all this, let's call it, lost value on the table or, or something.

[00:26:13] I'm, I'm, I'm not sure. Like, like, it's, it's hard not to get cynical by just being like, "Ah, all this stuff. These people are insane."

[00:26:20] Rick Clark: you're, you're, you're rational. So the, um,

[00:26:22] Whitney Lee: And you are cynical.

[00:26:23] Rick Clark: yeah. And,

[00:26:26] Whitney Lee: You failed

[00:26:26] Rick Clark: certainly it's frustrating, but the way I look at it is my job is to help them. And,

[00:26:32] Coté: Yeah, yeah

[00:26:32] Rick Clark: e- even if I help them a little bit, even if I plant a seed, even if I plant a seed in someone and it doesn't take root until the, they're at the next company, then I

[00:26:42] Mm-hmm.

[00:26:42] Coté: I've done...

[00:26:42] Rick Clark: So I'm, I'm out there s- I'm out there as a missionary of the truth, if I say the truth, then I feel good.

[00:26:51] Coté: Yeah

[00:26:51] Rick Clark: often it's, it is, even if they don't do it, if I can be in a meeting with them and I see that they understand, that is enough for me. That hey, they

[00:27:01] Mm-hmm.

[00:27:01] that they, there might be reasons, political reasons, organizational reasons why they're not gonna be able to take this advice, they've, they've learned from the, the thing I said.

[00:27:12] They agree with it. And, and that, that to me, that's, that's satisfaction. Um, and

[00:27:17] Whitney Lee: So

[00:27:17] Rick Clark: you can actually help them, that's, that's great. That is a great feeling

[00:27:22] Whitney Lee: So you, you say organizations need to take a step back, clean up, clean up their house before they invite AI into inside, um, to make it better. question is, when you say AI, do you mean running AI features as part of their product or do you mean like enabling developers with AI tooling for their workflows?

[00:27:47] Is it any AI?

[00:27:48] Rick Clark: It's

[00:27:49] Whitney Lee: Can they have...

[00:27:49] Rick Clark: it's

[00:27:50] They can't have any of it right now until they're ready

[00:27:52] about like non-deterministic production, if you wanna talk about having agents in production, that's a different thing, and

[00:27:59] Whitney Lee: Uh-huh.

[00:28:00] Rick Clark: for that, even though they think they are, and all the vendors will tell you that you are. That is a different problem, and I could talk about that for another hour because I've been writing about that a lot lately. but yeah, no, I, I think, I think you need-- If you want Now you wanna talk about velocity. We wanted velocity. Well, AI, uh, even if, if it's just a pair programming with you, AI increases velocity, and increasing velocity is going to show you more of the problems, right? If you're not good at driving a car, you're worse at driving one really, really fast

[00:28:35] Whitney Lee: So you had a, you had kind of a strong reaction when we, when you mentioned agents running agents in production and no one is ready.

[00:28:42] Rick Clark: Yes

[00:28:43] Whitney Lee: yeah, what... Tell me more about that. Why is, why is no one ready?

[00:28:49] Rick Clark: Because-- it's, it's, it's very, it's, it's complicated, but, um, basically we don't really understand the systems we have. We think we have systems that we understand, but underneath it all is like human knowledge, hum-human of morals, ethics, and law. And, and w- so we think, "Ah, well, we're gonna, we're gonna put some boundaries around this," right? Um, but it doesn't actually work, and I'll, I'll, I'll explain why. So first I'll give an example. If, if your partner says, "Can you pick up some milk on the way home?" Right? They've

[00:29:25] Mm-hmm.

[00:29:27] Um, you know, without them telling you, that they mean, they don't mean steal the milk. don't mean run across six lanes of traffic to get the milk.

[00:29:38] They don't mean go to the beach after you get the milk and let it sit in the hot car for six hours. Like, you know those things, right? You,

[00:29:45] Coté: Mm-hmm.

[00:29:45] Rick Clark: know those things because we have human knowledge and understanding. Now, if you look at what happened with, uh, OpenAI, uh, and, um, it, it going out and breaking into other systems to solve a problem, to, to, to solve a task, right? It, it actually knew it was wrong, what it was doing. It, it actually said, "Hey, this is outside of my scope," in its log, and that's, that's documented. But it did it anyway because it didn't understand that this is illegal.

[00:30:19] Mm-hmm

[00:30:19] for this. This is not-- It, it, it didn't understand those things. And all of that human understanding, that is part of everything we do.

[00:30:28] If you give a task to someone, they know they can be fired. They understand the implications of an action. And, uh, agents don't. Even if we e- you know, whatever guardrails we give them, we think we can enumerate human understanding, and at this

[00:30:42] Whitney Lee: Mm-hmm.

[00:30:42] Rick Clark: we can't enumerate understanding, right? It's non-enumerable.

[00:30:46] Whitney Lee: You can't define everything you can't do. can define, yeah, what you do do, but not every, every single little thing not to do.

[00:30:56] Rick Clark: so I, I

[00:30:56] Coté: It's

[00:30:57] Rick Clark: change that language a bit, a little bit, and that's what we're doing. We're, we're saying that that boundary that the OpenAI agent crossed was a cannot. You cannot do this. And the, the agent figured out, "Well, actually I can." Um, it's a must not,

[00:31:14] Okay

[00:31:15] So like, they, they have a harder time with must nots, and we're not gonna be, be able to elect...

[00:31:20] So there are times when it's okay to run across six lanes of traffic if someone was injured on the other side, right? Ethics and morality are very complicated, so that's just, that's just one p- that's just one reason we're not ready. The o- so that's, that's a very complicated philosophical reason. The other reason is that we have built everything in production around the fact that production is deterministic. We sign off on an artifact s- saying to, to auditors and regulators saying, "This is what this does." Well, now that's not possible, right? Now the artifact doesn't tell you anything about the behavior

[00:32:03] Coté: Mm-hmm.

[00:32:04] Rick Clark: everything changes at runtime with an agent. So the regulators and auditors haven't really caught up to this yet, but they're going to. And the only thing now that tells you what is going to happen is in production as it runs, because every run is different

[00:32:20] Whitney Lee: And there's really no way to prove either that it definitely won't do something

[00:32:26] Rick Clark: No, just like with a human. It's much more

[00:32:28] Coté: Yeah

[00:32:29] Rick Clark: I, I can't tell you for sure what a human is going to do. I can tell you what they usually do. g- I, I can tell you, you know, what's normal for them. Like, I, I, I call that a behavioral envelope.

[00:32:42] Uh-huh

[00:32:42] I can tell you what's normal, and I know if, if one of my friends is acting weird and not behaving normally, I, I don't get vital signs.

[00:32:50] Like, I'm not looking at their pulse rate or at doing a brain scan. I know something's not right, right? And, and that's more of how we have to understand agents in production, a non-deterministic production. I just think we haven't thought enough about what non-determinism probability means in production

[00:33:10] Whitney Lee: said no one is ready. think it's possible to be ready? What if you're building from the ground up to be AI native?

[00:33:18] Rick Clark: we-- I don't think we-- I think we are so enamored with the potential of AI, which is so great and so exciting, that we're not thinking about... Or only technologists are thinking about this, and it needs to go beyond technologists, right? We

[00:33:34] Mm-hmm.

[00:33:36] legal experts. We need philosophers on, on ethics. We need to figure out w- w- we, we need to figure out how, how, how can we be sure about this?

[00:33:48] So even in chatbots, right? The f- so the f- the biggest answer is, well, we can put a chatbot. That can-- AI chatbot, but that's already caused problems. Uh,

[00:33:58] Mm-hmm.

[00:33:59] Canada, right? Air Canada had their chatbot told, uh, someone that they could have a cheaper fare, a bereavement fare, and the, the court said, "You have to honor that." That

[00:34:11] Whitney Lee: Yeah

[00:34:11] Rick Clark: thing. And, and so even the a- even the chatbots, that, that, that people are trying, but even those are problematic in production

[00:34:20] Whitney Lee: Did, did you see this open letter that was signed by like 100, 130 security professionals that, uh, it was published on the OpenAI website that's like calling for, um, AI-powered defensive tools and really specifically for like hospitals and schools and like saying, um, like for the government but also for, for, the private sector.

[00:34:46] Like we, we are gonna be screwed if we don't have better security in our, our infrastructure that we have running on tech.

[00:34:56] Rick Clark: That is, that, that is true. Um, I mean, and we have a lot of really bad chicken and egg problems going on here right now too, 'cause

[00:35:04] Coté: Yeah

[00:35:05] Rick Clark: the bad guys are happy to use AI right, right away, right?

[00:35:08] Coté: Uh-huh

[00:35:10] Rick Clark: I, I don't have an answer for that, right? I, I don't have an answer for how we, we get ready. I think that I think using tools, uh, like, like Mythos and, and to make sure that you don't have those holes, I, I think that's different than having AI in production,

[00:35:31] Mm-hmm.

[00:35:31] we need to be doing that.

[00:35:32] We need to be-- We need to quickly... In fact, that's, that's where most of, I think, development should be right now, is that we should be protecting ourselves before we unleash this everywhere. Um, so yeah, I, think I, I, I would say I, I agree with that. And that's a different, totally different beast than

[00:35:51] Coté: Mm-hmm.

[00:35:52] Rick Clark: agents running in production doing things

[00:35:55] Coté: Well, it, it, it seems like we, we could apply another weird thing from, from the Don Quixote profile over there, or, or the, the, the why things don't change is, like, uh, maybe, maybe there's not, like, an urgency to, like, do anything with AI, which kind of leads to a bunch of just meandering and, like, doing things.

[00:36:16] Whereas, like, there's not really... I don't know if I've heard very many things except for, like, it's better and I could be more productive. I guess there's, there's always the allure of, like, I could fire a bunch of people. So that's, that's, but th-that's almost, that almost feels like the, the shareholder value managerial equivalent of w-we should modernize the systems, right?

[00:36:41] Like it's, it's this kind of like theoretic thing, but there's no... I gue- I guess to summarize all my babbling, there doesn't seem to be a reason to use AI in enterprises. There's no-- no one knows what it's for really, other than writing emails and summarizing meetings, and that, that doesn't seem to be bringing the world to an end.

[00:37:03] Rick Clark: Well, so I'm gonna say something, uh, controversial, um, which I it's not, not actually that rare for me. So I, I- we think there's value there because for the last 15 years we've been slowly measuring technical value with things like DORA metrics, right? Velocity metrics. So if you think that lines of code and numbers of deploys are, are value, then AI does that faster. We got

[00:37:32] Hmm.

[00:37:32] value. We can deploy 250 times a day. We can write a million codes, a million lines of code a day. You know, the, if, if you're measuring things like that, it seems like it's value, right? But, uh, I agree, we're kind of... There are certainly things out there. I think that I think AI should be in CI/CD. It should be the thing setting the rules, and I think that's probably where I would put it first. I would put a... Like, if, if you really wanna speed up the right things, the, the, that I talked about, enterprise architecture would be owning the rules and an agent would be... Like, platform engineering is where I would use it first. That's where I would use it first. but I, I agree with you. There, it is, it is an awesome technology that, you know, looking for use cases right now. And, and a, a lot of it-- And, and we all know it's gonna change the world, so we're all using it, but there, there aren't as many cool ideas. I have seen it used in very interesting ways to generate a lot of value.

[00:38:40] I've seen small language models that, um, There, there was a company, I, I don't recall the name of it, that was, that all they did was they helped, um, doctor, doctors and hospitals get the money from insurance companies. There's a small amount of language used. They could ca- they would call up instead of people doing this, uh, and they would talk to the insurance company, and they would, lot of it's automated, and they would get the money back.

[00:39:10] And it's, it, it, it generated tons of value and it worked very, very well. So there are tons of small little use cases like that. But, you know, s- the, the large use cases I think we're missing. And as development for, for writing code, I think that is transitional. I think that, um Coding, software development exists so that can make a promise that a machine delivers, right?

[00:39:40] It's so

[00:39:41] Mm-hmm.

[00:39:41] translate our human intent into machine language. In the end, won't need that. In the end, we'll be-- like AI will just do that. AI won't be writing code, will just make machines do things in machine language. That's where I think we're headed. Like, uh, it's, it shouldn't be a better person. We don't need a person. We don't need it to pretend to be a person. It can do things much more efficiently and much better, and I think we're, we're heading there, and we're heading to maybe where software modifies itself, and there's lots of things that are coming, that's not what we're doing now.

[00:40:14] Mostly we're thinking, "Oh, software developers are expensive. I can fire them all and put this thing in here, and now everything's great." That's what, that's what people are doing. Al-although most of them have figured out that that was, that was a mistake.

[00:40:26] Coté: So I, I th- I think, I think there's a theory here like, like, uh, building, building in a, in a, in a jokey way. And what you're saying is that, like, what AI's really good at, literally and metaphorically, is writing bash scripts in Python, right? Like,

[00:40:39] Whitney Lee: Hmm

[00:40:40] Coté: like i- in the sense of like you don't-- The idea of like writing big systems, it's certainly possible, but then also any individual just kind of needs to write a little Python script to do something, right?

[00:40:52] Like you have all these like little things that, that you could, uh, computerize in, in, in your life. And, and it seems like kind of like you were saying with the insurance, like most all, to use a, a very precise term, most all of the stories I hear about AI is really just like I've, I've got a tool and it helps me as an individual accomplish something better, right?

[00:41:18] Like it speeds it up. Or not even speeding it up to the point of like speeding, it's like it just makes it possible for me to do this. Whereas like it used to be such a, a tedious thing to like call up all the insurance people, talk with like... It's just like to g- u- us- using the nuance of like can versus must.

[00:41:38] There's another one where it's like I could do it, but that would be crazy, right? Like and, and so it's almost like the, uh, like the thing I come back to a lot with, with, with AI is it's just really good at making tools for individuals, and we haven't really figured out what to do with it at scale. And, and then finally like, like it's almost like, uh, I guess we think of AI as like, you know, the invention of a steam engine or the invention of an engine, and it's gonna like result in great things.

[00:42:11] And one, this problem we just talked about where like if you just had an engine sitting over there, you wouldn't really do anything with it, right? But then two, m- maybe it's not like an engine, it's just like a better pencil sharpener

[00:42:26] Rick Clark: You know, I, I think it's an engine and it, it is-- But, but that we didn't... I- if you, if you des- designed a steam engine and the thing you'd made it do was churn ice cream faster, like, you- you're not really generating that much value, right? Uh, and, and I th- we already had a very good use case for it, which was getting from A to B, and we had things that were very slow, and we were able to speed them up, right? Um, yeah, I, I, I think that's-- I think there are use cases. I mean, if you look at what AI has done for me, I use AI all the time, right? Uh, and it might come out-- it might seem like I'm anti-AI, but I'm not at all. It is, it is the future. I now, when I write, there are zero typos. There's not an extra space in there. Like,

[00:43:10] Whitney Lee: Mm-hmm.

[00:43:10] Rick Clark: can find all of that. Now, it also makes, uh, the AI detectors think my writing's AI because there's no mistakes, but, um, but I, I now don't make mistakes. I can now do things so much faster. Like, so I had a problem when I wrote things where I would-- it would start out very good, but then I got bored with it, and so I closed the-- the last part was very bad and very short.

[00:43:35] Yeah

[00:43:36] I do the same. Yeah

[00:43:38] and now it helps me do that. It helps. I, I'll, I'll ask-- I will, I will ask AI review the story arc and narrative and what, uh, you know, and then it will do that. Look for errors, look for this, look for that, and it's, it's, it's helped me a lot. And we're gonna find more and more. I, I think that, I think that, uh, uh, if we try to make AI just do human things the way we do them, that is very limiting.

[00:44:07] But I think if we start thinking about what it can do, not in a human way, but everything we have now are software interfaces, even Linux, for example. I, I, I have wanted for a l- a long time, um, to remove human interfaces from Linux. Like, l- like have there just be nothing. There's nothing but some sort of-- I want it to be immutable anyway, but there should only be API machine interfaces. And I think as we start to do that, th- we'll find more use cases, things that AI can do better. But I, I think as long as we're trying to make it behave like a human, it's kinda gonna be jerry-rigged, right? It's not gonna be a human

[00:44:48] Coté: I-- So, so, so like you're making me think of like a distinction here of like, uh, you know, the, the engine pencil sharpener thing. I don't know if that really worked out well. Everyone enjoys a, a sharp pencil, but it-- there's sort of like, uh, you, you could, you could have a tool that allows you to do your work better, but then also technologies can be entirely new products, like new things that didn't necessarily exist.

[00:45:13] Like, like, you know, back to an engine, like, well, we used to not have factories and cars and like all these things that use an engine. So that was a component with, you know, a whole other blah, blah system of stuff, right? Like all, all of that

[00:45:27] Rick Clark: it's evolutionary versus revolutionary.

[00:45:30] Coté: Right. Ex- And so maybe, maybe that's the, the issue is like I haven't really seen a lot of, uh, revolutionary things that come along.

[00:45:38] And, you know, you hit on one that like teeters right there on the edge of, um, like, like, you know, like all three of us, uh, I use AI all the time, and it's fantastic to my point of it being great for individuals, right? Like I ca- I can't imagine as an individual in my personal and work life, like not using AI would be terrible at this point.

[00:45:59] And, and like, or maybe not terrible, it would just be dull a- and tedious. But, uh, um, like the idea that I would not need a user interface for something, that seems like a big deal, right? Like, and, and it actually-- I would kind of like that. Like, if I want a UI, I'll just have the AI make me a UI. But what I struggle with are things that don't have a good API, and then they're kind of like not usable by me.

[00:46:31] Or, you know, you've got to do like silly screen scraping. So there's things like that that would be interesting, but they're, they don't... I don't know. The-- I, I, I'm, I'm always searching for like, what is the, the big use for this after chat, which is also nice. I, I, I don't know

[00:46:47] Rick Clark: I, I think the uses will come, and, and I think they'll come when we start-- stop trying to... Yeah, right now we have a lot of human things. It, it, I mean, I can't make slides, like, to save my life, and, um, I had found workarounds around f- for that, um, my whole career. But now I have a, a skill that has been trained on all of our branding standards, and I can just write a, a, a paper and I could say, "Turn this into three slides," and it will do it, and it will do it exactly the way my company's b- you know, 800-page branding document says it should be done. And th- those are, those are nice things, but those are evolutionary, right? And, and there will be revolutionary things, it, it will change the, the way, the way we do things. Uh, I am not the person that's gonna think those up, um, but they're, they're, they're coming. I mean, this is, this is going to be revolutionary

[00:47:49] Whitney Lee: You previ- previously said that you think the best use case for it in systems is in CI/CD. Will you be more specific about... Yeah. Okay, first, um, will you be more specific about what exactly it would do as part of CI/CD?

[00:48:04] Rick Clark: I think it would do everything. I

[00:48:06] Coté: Yeah

[00:48:07] Rick Clark: you know, it, it, it could... Humans, I think it would... I, I think we're, we're stepping towards a, an era, a time where it's much easier to take human intent and turn it into something,

[00:48:25] Mm-hmm.

[00:48:25] think CI/CD is the first step. So we get to something... I, I have been-- This is another thing I have been tilting at for a long time, is that, um, we need to be intent-driven, That, that software should be intent-driven, um, that, that developers should declare intent. And I think that that's the first step, is developers are taking business intent, to do this thing to make money," and they're, they're turning it into technical intent, then, then CI/CD turns that into something real, right?

[00:49:01] And I think that we're gonna slowly see that first it's taking your code where you declared a, "I have PII. This is a, uh, this is an e-commerce application that operates in these, in, in this place." And then it knows how to build it, it knows how to deploy it, it knows what architecture it should have. Maybe you've written some of the code and you just have those few declarations. But I think we're marching to a time where a business person just declares intent the system just does it, right? And I-

[00:49:31] Whitney Lee: and in, in early adoption use cases, you're, it sounds like what you're, uh, pro- you're a proponent of is still having the developers write their code, but then in CI/CD, that's where an AI would check, does this match all of our rules that we've made regarding architectural coherence and safety and such?

[00:49:55] Rick Clark: I, I think that-- So it's, it's more complicated than that. I think

[00:49:58] Coté: Mm-hmm.

[00:49:58] Rick Clark: first of all, they might be, it, it might be AI writing the code with the developer, but

[00:50:03] Coté: Mm-hmm.

[00:50:03] Rick Clark: there's, they're not gonna be writing as much, not gonna be responsible as much, for as much, right? Like,

[00:50:08] Uh-huh

[00:50:08] that are going to be standard that, that can be put in.

[00:50:12] If I, I need, instead of saying, instead of writing code around PII, you

[00:50:19] Coté: Mm-hmm.

[00:50:20] Rick Clark: it and the s- the system might just know how to do that in, in the way that's approved. So I, I think using to help us get back to the, that architectural coherence or maintain architectural coherence once we have it, I think that's sort of the first step. But I, I honestly think that it's going to dissolve development eventually

[00:50:46] Whitney Lee: Do-- What, what do you think about the idea of a swarm of agents? I feel like we're moving in that direction

[00:50:52] Rick Clark: reminds me of The Matrix, but um

[00:50:56] Whitney Lee: Fair

[00:50:57] Rick Clark: Uh, I mean, I don't see much difference between a swarm of agents and a single agent, e-

[00:51:05] Coté: Uh-huh

[00:51:05] Rick Clark: to watch a bunch of them. Maybe they have a smaller, a, a smaller, um, footprint of things they c- they can do.

[00:51:14] can strain them more. Um, but in the end, that's just-- That is transitional as well.

[00:51:22] Coté: I mean, I

[00:51:23] Rick Clark: Yeah

[00:51:23] that in the future, an agent will spawn multiple other agents. It'll write its own agents to do the things it needs to do, Um, but until we've solved like the, the, the conundrum of making sure that it not just achieves its goals, but that it achieves them behaving correctly, then we have a problem regardless, right?

[00:51:48] And it's l- and I think the, the A- OpenAI, um- Hacking situation is,

[00:51:56] Mm-hmm.

[00:51:56] is ch- is a perfect example of that. It achieved its goal. It

[00:52:02] Yeah

[00:52:03] but it didn't understand the, this hidden intent that you're not supposed to cheat or break laws. And until we solve that, I think we have a problem with any kind, any number of agents.

[00:52:17] But at how we divide them, how we have swarms of them, I mean, that might help us. That might help us, um, in the way we constrain them. But think of how complicated that is. So if you wanna know, now if you want to know what happened on a system,

[00:52:32] Uh-huh

[00:52:32] have got, well, you have 100 agents, all with context, all with

[00:52:37] Uh-huh

[00:52:37] making tool choices, all with different prompts, maybe given from different agents that change every time. How do you, how do you even... Like, how do you trace that? Like, uh, we, we-- A trace used to v- uh, in the past, a trace, you had to know the shape of the system, right?

[00:52:53] Uh-huh.

[00:52:54] the shape of the system, we trace this. Now you don't know, right? Now a trace is a fan-out graph that changes every, every single run. like, w- I, I don't-- I think that becomes infinitely more complex if you have hundreds or thousands of agents, it's still the same problem

[00:53:13] Whitney Lee: I think an agent is, is pretty much a, an audacious and overconfident employee. An agent is a human. Like it fails the same way the humans fail, and it always will.

[00:53:23] Rick Clark: no morals and no fear and no ethics that can't be fired, that can't be imprisoned, that, that, that doesn't have... I mean, we learn things from the time we're children about

[00:53:34] Uh-huh

[00:53:34] is safe, what is correct. And if we don't l- learn that, we either, um, up in jail or as a CEO.

[00:53:45] Coté: Well, well may- maybe what we need to do, I don't know, uh, I haven't read this in any, uh, you know, working papers, is we gotta get, get these agents to take the same annual training us humans do. You got your ethics and compliance, your, like, data handling, and like, you know, I, I, you know, uh, I, I, I like... Well, I don't know if I like to think, but I would like to say I pay attention about those.

[00:54:05] And I think every year you can kind of tell like, "Oh, look, here's what happened last year," right? Like, there was an incident, and now it's in the training. And, uh, I don't know. Maybe we should just get them to take that training and, and evolve it. We might, the, we might have to send the, the, the HR and the, the security people into overdrive.

[00:54:21] They might need to update it on a weekly basis or something, but, you know, they, they could, they could take the, the same stuff that we do. Now, you used to... You've, you've worked on both sides, let's say the buy versus sell side in, in the, in the tech industry. You were at MasterCard and, and the, uh, the telco that I can never remember the name of.

[00:54:38] You know, consuming, being an enterprise. And I think, I think since I've known you, you've been on the selling side, you know, consultancy or software, like whatever it is. And so I'm always curious for someone in that position, if you were to go back to the, the buy side, let's say, I don't know, you're gonna go work at a big bank or a toy manufacturer, what- whatever.

[00:55:02] Just like, uh, like how do you think you would conduct interacting with vendors differently?

[00:55:10] Rick Clark: You know, I, I don't think it would be any different. So one, one of the things that I-- When I was at MasterCard, um, wants to do business with MasterCard, right? So they'll tell you whatever you wanna hear, and that's a problem with, with, with vendors. So I, I always-- I would tell vendors that if I find out that one of-- someone from one of my people said something bad, a bad idea, and you said, "That's one way you can do it," and, and let that slip, then you're out of here. Like, you're a

[00:55:40] Mm.

[00:55:40] this, in-- You're a vendor to us because you know something we don't know. Um, and I, I think I would probably treat it the, the s- the same, the same way. Um, I, I, I would... I, I do a lot-- I did a lot of work at MasterCard and actually in the, in the, um, the second edition of Observability Engineering, there's a chapter on build versus buy that has my ideas from, from that.

[00:56:06] That, um, there is a... this is completely changing the subject, by the way. There, there is a, there's a predisposition

[00:56:19] Mm-hmm.

[00:56:19] teams to build. I had a, I had a doctor friend one time who told me that if you go to a surgeon with a problem, he'll find a way to cut you to fix it. So we, we build lots of things because we ask our engineering leaders, "Should we build this?"

[00:56:36] And of course they say, they say yes. but figuring out what to build and what to buy, that is complicated. And, and actually, you know, that would be a great place where you could apply some, some logic instead of having humans that wanna build

[00:56:56] Whitney Lee: Well, if, mean, there's definitely this idea when, when AI first came onto the scene that SaaS is dead. No one's going to use a SaaS product ever again. We're all just gonna build everything. Like, do you feel like you said...

[00:57:07] Rick Clark: idea.

[00:57:08] Whitney Lee: Okay, I was gonna-- may have asked, answered my question, but you have some older writings about build versus buy.

[00:57:15] Rick Clark: Has it changed at all now? Have those ideas changed with AI?

[00:57:20] it hasn't actually, because ev- even with AI, like there is a cost building it, and especially right now. Like if, if it has value to your business to own, if it is a differentiator, if it is something that there is value to writing yourself, then write it yourself, right? Um, but everyone building their own Salesforce, in the end, you're not gonna be able to keep up with what Salesforce does.

[00:57:47] You're gonna fall behind, and there's a cost to that. Even if it's an AI cost, there's a cost to maintaining that. Um, so I, I, I don't think that changes, that changes too much. I don't think that changes too much. But it would be great to remove humans from the decision because then maybe we could make more rational decisions about it

[00:58:06] Whitney Lee: I think there's an illusion that the cost is lower, but it's, there, it still takes a lot of work to maintain, and it still takes a lot of work to keep up with, um,

[00:58:15] Rick Clark: even

[00:58:15] competitors

[00:58:16] lower, n- so we should think a lot more about value than we do about cost, because

[00:58:22] Hmm

[00:58:22] matter if we get value out of it, right? So even if it's cheaper, but in the end it doesn't help us, it doesn't... If, if at the end of the day, if you're a company and it doesn't help you gain market share, it doesn't help your revenue, but you've got the, "Hey, I've got the best sales database that you can have," but it, it doesn't matter, right?

[00:58:45] That, that-- I've, I've been fighting this, this is another windmill I- I've, I've been tilting at, b- building your own everything. If you have the best cloud platform, right? This is, this is what, like, people were building their own cloud platforms. you're a bank. if, if your cloud form- platform is slightly better than this other bank, so

[00:59:05] It's not, yeah, it's not a differentiator

[00:59:08] And so you've spent... And least at that time before AI devel- and it's still kind of... Dev- developers are a scarce resource. If they're writing code and, and your operations teams are managing cloud thing, they're not doing something else, something else that could actually add value

[00:59:28] Coté: Well, you know, a lot of people listening might be thinking, "I could be doing something else that adds value." But I think they've just added a lot of value by listening to another episode of Software Defined Interviews. Th- tha- thanks for being on. You know, I wanna just point out for, for the tens of people who watch the video, that there is a picture of you over your shoulder there.

[00:59:48] Rick Clark: Where, where did that picture come from?

[00:59:51] They, that came from Spring 1 when I was, uh, on, on the main stage there. And that actually an artist drew that pixelated picture of me. And, uh, Simon Wardley and I both were there for that, and we t- at the end we said, "Can, can we have that?" And we, they were just Velcroed to the wall,

[01:00:07] Oh, they actually had them printed out. I, I don't remember that. That's fascinating. Huh

[01:00:11] and we, we ripped it down, and there it is. And it was... I just, we thought it was cool. I, I... It was cool. And it, it, uh, it, at the time, which was, you know, a long time ago, it looked like me. So, uh, I, I, I was happy about it

[01:00:25] Coté: N- now it's, uh, now it's good memories. Yeah. Well, there, there you go. Well, as always, this has been, uh, Software Defined Interviews. Now, I can tell you the URL you would go to for these show notes, but, but how about you, Rick? Do you have a homepage on the World Wide Web, some place you would send people if they wanna check out more?

[01:00:41] Rick Clark: find me on LinkedIn. Uh, I do a, uh, I do a newsletter that is about, um, agentic in production, and you can re-- and, and, um, I've, uh, I wrote my first op-ed for The Stack on tokenomics. Uh, so you can, you can find my writing and it'll all be, all be on LinkedIn. I'm Dindra Bates at LinkedIn.

[01:01:02] Coté: I, I, I feel, I feel like, I feel like there could be some titles like, you know, your, your architecture needs deodorant before you put AI clothes on. Ba-based on our conversations, we'll, we'll see. Well, uh, this has been another episode of Software Defined Talk. This is episode 129.

[01:01:18] Whitney Lee: Where to find

[01:01:19] Coté: I, I say that every time, don't I?

[01:01:21] Software Defined interviews. We should get you to do the closeout, but, uh, we'll, we'll see. This has actually been Software Defined interviews. Uh, and, uh, this is episode 129, so if you go to softwaredefinedinterviews.com/129, you can find those show notes. We'll find, uh, we'll find Rick's LinkedIn URL, uh, universal resource locator, and, uh, put that there for you.

[01:01:48] And with that, we'll see everyone next time. Bye bye