Stop Ripping Out Your Legacy Code with Alan Williamson, CTO

(Alan at 00:00:00) The 10x developer is just nonsense. 10x what? I mean, the amount of code you generate, the amount of time—it's just a silly metric, and particularly in my space where we are sort of putting lipstick on the pig in order to sell it to the next organization. But when you sort of scratch the surface, you think, you know, that's just a cron tab? There's nothing AI here whatsoever.

(Alan at 00:00:23) Let's not just go after the shiny new toy because it looks cool.

(Joel Beasley at 00:00:33) Alan, it's been a while, hasn't it?

(Alan at 00:00:34) It has been nearly a couple years now, if not a little bit longer.

(Joel Beasley at 00:00:38) Yeah. For people who didn't listen to the first episode years ago, tell me a little bit about your background.

(Alan at 00:00:44) Yes. So I am a professional CTO working mostly in the private equity space. The private space I work in is for companies coming into the private equity space for the first time. So it's professionalizing their tech team, getting them all sort of set up, and effectively professionalizing them so the company can grow.

(Joel Beasley at 00:01:04) You're a PE firm that helps PE firms become PE firms?

(Alan at 00:01:09) Well, no, no. We're helping the companies that PE buys.

(Joel Beasley at 00:01:14) Oh, you're helping the companies that they're going to acquire.

(Alan at 00:01:16) Yes, yes. So both due diligence, and then it's looking at what can we do to help grow that company, stabilize the product, get it into more compliance, and effectively scale it.

(Alan at 00:01:27) Because most of the time, you get yourself into a situation in these founder-led companies where it's that one guy that does everything. It's one guy that starts everything, and when they go on holiday for two weeks, pretty much the tech team stops for a while. It's getting that company scaled beyond that one person so that we can grow the team, we can accelerate the product development.

(Joel Beasley at 00:01:48) And you're also an author? I think we've had you on about the book last time.

(Alan at 00:01:51) Yes, I did. The Think Like a CTO book has gone phenomenally well. In fact, since we last spoke, it got republished in two different languages—both the Japanese version, which is phenomenal. I love the feel of it. It feels like a manga book from that perspective. And then it got translated into Russian. So I'm big in Russia for whatever reason.

(Joel Beasley at 00:02:17) At least that.

(Alan at 00:02:18) It's been kind of cool. Yeah, that's excellent.

(Joel Beasley at 00:02:23) All right. So the book's going well. Where could people buy the book? Just to get that out.

(Alan at 00:02:27) Amazon right now, the usual stuff. Or you go to my website, alan.is. You will find a link there to get you to where you need to go in three different languages.

(Joel Beasley at 00:02:34) I feel like my book—I wrote it seven years ago—I feel like it's outdated.

(Alan at 00:02:39) Which bit do you think is outdated? Because I looked through mine, and outside of some of this AI stuff, it's still holding very strong.

(Joel Beasley at 00:02:48) Yeah. You know, I think that they were all principle-based chapters. So I think if you dig the principle out of it, those still hold. It's just like references, and I guess the biggest thing is what I'm not addressing. I think that's what bothers—I think it's less about, is that content, do I stand by it still? Yeah, I stand by it still. But I've learned so much and I've seen so much happen that I wish I had the desire to sit down and write a version two.

(Alan at 00:03:17) I think the chapter, Joel, that has really surprised me—because it was more of a, I wouldn't say a filler chapter, but it was one of those chapters where I need something nice to end the book with—and it was the imposter syndrome chapter. And I spent a bit of time on the imposter syndrome chapter, but that is the one, in this day and age of AI, that has really resonated with a lot of people and have come back. And the amount of feedback I've got because of that chapter—because I basically outline that it's okay not to know everything. It's okay to be a little bit behind the curve. Don't panic. The world will still continue to rotate. That one has helped a lot of people sort of get their arms around this fast-moving AI space. And we are in such a fast-moving space that the standards are still being evolved, and what was working last week and a new subsequent release of new software suddenly stops working because the models are changing, et cetera. I mean, we've never been in such a rapid changing environment at this precise moment. And when you're a CTO trying to plan out the next two to three years of technology and stack and trying to figure out, okay, where are we betting the farm against? This is a very difficult time to be in that leadership position.

(Joel Beasley at 00:04:31) And so you feel like—because the book only came out like a year or two ago, right?

(Alan at 00:04:34) Yes, but it was pre the start of the AI hype, in all fairness. I think if it had waited another six months, I would have caught it.

(Joel Beasley at 00:04:42) It's hard to keep the conversations, you know, as we do this content stuff—it's hard to keep them from AI. AI is in every conversation because it is this new technology that is affecting everyone. The guys who do my lawn use AI to help calculate their schedules. They use it to take pictures and design the plant beds with the customer. An 80-year-old woman in my neighborhood uses Gemini, and she took a picture of the backyard, had the rocks changed, and she showed it to the landscaper. She's like, I want it like that. I'm like, you're 80. This is amazing.

(Alan at 00:05:16) It is, it is. And I think as we're—I mean, I personally think we're sort of coming off of the golden age of AI at this precise moment in terms of accessibility to all of the models and the cost of the models. There's no way our $20 a month subscription is going to fund what the AI industry needs it to fund. I mean, it's a trillion dollar business. They've got a huge check to write, and the economics just don't work. So I think, as engineering teams sort of wrestle with that, token costs are starting to become a real factor in how we manage this sort of stuff. And I think, between that and—I also think data sovereignty has become a hot topic as well, which is the AI companies have already been caught with their hand in the cookie jar with respect to, no, we're not training off of your data. Well, you are. And certain people have stood up and said, yeah, yeah, we are. So we're now looking more towards local models and hosting the models ourselves far more. And I've been involved in a number of engagements on that front, and it's been surprising to sort of lift the lid on that for people, to show that it's not as far-fetched or as complicated as people may realize it is in order to—and it's not necessarily reducing the cost of the AI. It's just having stability and predictability in and around that infrastructure. If you're going to be building AI into your product, then you need to be able to guarantee that level of service even though OpenAI's API has just gone down or Anthropic's API has just gone down. And they do go up and down very regularly. So from that perspective, local AI has started to become a real alternative to the big boys.

(Joel Beasley at 00:07:18) You mentioned something interesting that came up on a show the other day, and I just want to clarify to make sure I understand correctly your thought on this. You said the $20 subscription is not going to cover the economics for the trillion dollar check that they're having to write to build this technology. Now I talked to somebody—and I can't remember exactly who—but a couple weeks ago. And I'm not sure if it was in the recording or after, because we have fun conversations about it. But it was something along the lines of, from an investment perspective, they're worried because they don't think these companies are going to be able to go public easily because they're spending far more than they're charging for us to do this. And when that hits the public markets and everyone sees that, he thinks there's going to be a massive pullback, because if you're operating at this huge loss, so you're forced to go public and operate out of loss, everyone else sees you're swimming without pants on. And then what happens to the cost to us, but also the loss of value of NVIDIAs and other technology companies that are in the ecosystem. Is that what you're kind of talking about?

(Alan at 00:08:24) Yeah, absolutely. And I think, as early investors are hoping that the new investors are coming in to pay off the older investors—I mean, it's a pyramid scheme by another name.

(Joel Beasley at 00:08:34) That's exactly what he said.

(Alan at 00:08:36) Yeah, yeah. So it's a bizarre situation. And I think every service that has got AI involved in it, you're running up tokens somewhere. And if you've got a product that—and again, it's that sort of, does your customer really care that it's AI-driven, or are they more caring about that the product is solving the problem that they're going after?

(Joel Beasley at 00:09:02) The second one.

(Alan at 00:09:03) Always. But you look at how many products are out there, or SaaS-based products out there, that have got that sort of token usage as part of your plan. You can use X number of images or X number of documents per month before you have to pay more. So already you're showing your hand in terms of your cost of delivery of that particular product is—okay, the real cost is underlying AI, not the actual SaaS-based product to which we're doing. And again, what's been interesting for some of the engagements that I've been involved in—people want to use AI sometimes for the sake that they think they need to use AI. But when you sort of scratch the surface, you think, you know, that's just a cron tab. There's nothing AI here whatsoever. Yes, we could use AI, but that's just a script. So pull off of that, and let's not just go after the shiny new toy because it looks cool.

(Joel Beasley at 00:10:01) Well, it's like we have the colors of the rainbow, right? Like our color spectrum. It's like we got a new color. Everybody just wants to be like, ignore all the other colors. I'm using this one. This is the answer for everything. It's like, no, no, no. That's an important new color, and it needs to be used correctly, but it's not like all the other colors don't exist.

(Alan at 00:10:22) No, that's a very good analogy, Joel.

(Joel Beasley at 00:10:25) Well, I just made it up just now.

(Alan at 00:10:27) Well, I mean, yeah, I was here.

(Joel Beasley at 00:10:28) I see it. You inspired it. I think yesterday we used a similar one about like a salad bar. Like they all play their different roles. The cucumber needs to be the best cucumber it could be. Today we're using color spectrum. We're going to keep going.

(Alan at 00:10:44) And I think sometimes—like there was one engagement we were involved in where they had basically layered on the whole AI marketing thing a little bit too much onto the product, that AI was powering this. And long story short, the customers just rebelled against it and said, what? Stop. Wait a minute. What happened was that it made the customer feel irrelevant. It sort of patronized the customer and sort of had to pull back from that. A few months later, we did the exact same thing, but instead, this time, we called it "enhanced." Removed the word AI completely, and everybody loved it. So again, I think we're coming off of that sort of wave of where we have to put the word AI into everything. But now focus on, well, what are we getting for it? Why am I coming to you as a service provider as opposed to you telling me, well, we use AI? So what? Are you solving my problem?

(Joel Beasley at 00:11:46) Yeah. I use AI. I just wake up. I haven't planned my day. So my whole company, therefore, uses AI. AI is in everything. Wait. Question for you. You've mentioned this a couple times—people don't like it. Where was that specific situation geographically?

(Alan at 00:12:00) The US.

(Joel Beasley at 00:12:02) The US. Okay. So US consumers were actually pushing back on—and because this is something I find fascinating, there's different age groups, different experience, different income levels, and they all kind of see AI a little bit differently. Obviously on the spectrum of love and hate, right? Like I love it, I hate it. But the thing is, I'm not finding—usually when this type of stuff happens, like I can kind of sense it amongst my social circles as I'm out and about, the people I talk to. I can't peg anything on this one. The moment I see a type of person who I'm like, that person's going to love AI because all the people I know that are like them, I meet someone who hates it that's just like them. And I can't find out like what this common thread is.

(Alan at 00:12:47) Yeah, it's an interesting one. I couldn't put my finger on the demographic or even the social background of them. What I would say is that I deal a lot with non-technical companies that are using technology to deliver their product, as opposed to the product being technology. So therefore, they're not in the same echo chamber as to what most technical people are. So they're seeing it from the data centers coming into this sort of stuff. They're seeing it from creative people now being sort of put out of jobs, and also the mass layoffs that are being put on the AI doorstep of blame—which I would argue that most of the time, it really isn't AI's problem. It's probably an over-hire in the COVID space that companies are getting rid of an awful lot of people. But it's convenient to blame it on AI from that perspective. So all of that sort of mixing into the whole discontentment with what's going out there. And even, you know, this week, Anthropic has said they're going to start watermarking the text that are being generated so we can see what is AI slop and what's not. I mean, LinkedIn has now got—they think this is it—you need the ability to report the fact that this post is now potentially slop, to get the wisdom of the crowd to sort of identify all of this. And I think that's an interesting trend, which is, okay, let's stop pretending, but let's be more honest as to what is AI-generated and allow the consumer to make the decision therein. So if that particular website—be it news, be it whatever—is 90% AI-generated and people no longer visit it, then, okay, there's your answer. It's not working. But if people still consume it, then fine. But I just find it interesting now that this technology—and most technologies do stay within the sort of the tech sphere. You know, crypto, big data, Web 2.0. These are all things that have always stayed within our world and has never spilled out over. You never read about it on CNN or BBC or any sort of mainstream media for that. AI has escaped the farm, and everybody is seeing it from that perspective.

(Joel Beasley at 00:15:13) Yeah. You can be at a restaurant, and 15-year-old kids are talking about it in the booth next to you. It's—everyone is engaged with it. I think—and it's funny because everybody uses the same word. You know what it reminds me of? Reminds me of God. It's one word, but everybody sees it so differently, and it's so emotionally charged based off of who's bringing it up in what context and—

(Alan at 00:15:37) Joel, you're on fire today. I mean, that's art and color. I mean, it's God. So we just need one more.

(Joel Beasley at 00:15:43) It's not happening, Alan. No. But isn't that how it is? Because you bring up AI, people go off the rails, or they love it. But then you and I are practitioners, so we know. Like we could actually look at the difference between the large language model type of response, deterministic versus non-deterministic, and we can have those conversations and which tool do we use when. And if you say you don't like AI, well, if I edit clips from this video, we have an AI tool that identifies the most potentially viewable clips and—

(Alan at 00:16:17) Gotta be the color one. The color one's going to make it. Yeah, it's going to make it.

(Joel Beasley at 00:16:20) But is that AI generated? It becomes like a percentage game. It's like, oh, this content was 38% AI generated. You know, because it takes...

(Alan at 00:16:28) That's a good example, Joel. I would say that that is AI being used as a tool, but it's not generating the content. AI isn't gonna generate what you and I just responded to each other there. But for AI to come in as a sort of like a virtual director and say, I think this segment is actually gonna hit and resonate as an Instagram clip or as a LinkedIn post, etcetera, for 22 seconds versus making people listen to the 45-minute podcast. That is wonderful use of AI. I mean, it's the classic, do I wanna sit and look back over all my security cameras to find out who walked into my yard all night? No. That's a perfect use of AI. Just go tell me the hour that I need to go and look at, and I will dive in deeper on that one.

(Alan at 00:17:07) Which, again, brings us on to an interesting one. We're seeing that a lot more in the diligence world. So when we buy a company, the technical due diligence has effectively moved along. Where technical due diligence was, you would get to look at all of this massive source code and database tables and architecture and all of that stuff, and the logistics was you could never view everything by hand. You just had to sort of gut feel pick at certain areas. I can, I better look at the security, I better look at the backups, I better look at blah blah blah. AI has now got to the point where, yes, I can run a handful of agents across the whole ecosystem and give me a good report. But the real benefit of the AI diligence report now is flagging those little flashing spots that require a human or somebody like myself that knows what they're looking for to dive deeper into that area. Okay, so it has found something here. I don't need to look at the whole 24 hours of camera footage. It's now told me a quarter past two, something happened. You better zoom in on that part. And that's where AI is really starting to kick.

(Joel Beasley at 00:18:23) Yeah, where it's identifying moments we need to examine.

(Alan at 00:18:26) Yes. And it's not taking the human out of the loop. Right. And that's the important part.

(Joel Beasley at 00:18:34) You're right. And do you think that in the future, that content, because it's such a hard conversation to have and like, where do you draw the line at AI generated? Like, if it generated lower thirds. I almost think that there's this weird future. I hope it doesn't happen, but I almost think it could where you could click on, like, the nutrition facts to see what, like, this video was 38% AI or whatever. Like, do you think that's gonna happen?

(Alan at 00:19:05) I did read an article the other day where Wikipedia is desperately fighting the amount of AI generated content that's being added, and the editors are really fighting against that because it's not being fact-checked from that perspective. So if there was any site that would need that particular warning, Wikipedia would indeed be that one.

(Joel Beasley at 00:19:27) Mhmm. You know what's been frustrating for me? Upwork. I love Upwork. I don't know if I can even include this. I think you could say if you don't like something about them. But the problem is, I used, I've been on Upwork for over 10 years, like, hiring people. And you used to be able to tell how good the person was and how well they understood the problem by their response and application. Did they actually read this? But they've all hooked up their AIs to automatically do it. And now, my response to everyone is, okay, can you actually take a look at this? I don't even respond to them. I see their response. I'm like, okay, can you actually take a look at this and let me know if you can do this project?

(Alan at 00:20:09) Wow. Yeah. Yeah. Yeah. And I, well, I think that's the counter argument to everything, is that AI, whether it's emails and internal document, Wiki, Confluence article, we seem to be producing more content. PRs are bigger. Jira tickets are bigger. Shouldn't it be going the other way where AI is generating less content for us to read and really summarizing this? But we seem to be enjoying this 20-page document I can now produce. No. I don't wanna read it. Thank you. Could you summarize that into like two or three paragraphs and tell me what the pertinent information is? So, again, I think we're generally coming off of that AI as that novelty tool. Yes. You can produce a book. You shouldn't, but you could.

(Joel Beasley at 00:21:08) Have you seen, there's a meme out there that I saw recently and it's got, um, on one side, it's got, oh, hold on a second. I actually just found it. Look at this. I'm gonna share my screen here. It says, look at the left. It says AI turns this, it's a person sitting down at a desk and they're talking to their coworker and they're like, look at my computer screen. AI turns this single bullet point into a long email I can pretend I wrote. And then on the flip side, the manager is talking to their friend, and they're saying, hey, look, AI makes a single bullet point out of this long email I can pretend to read.

(Alan at 00:21:42) Yeah. Nailed it.

(Joel Beasley at 00:21:43) Nailed it. So good.

(Alan at 00:21:47) Yeah. There was a great phrase I heard, and I actually found it on a Reddit forum, and I'm bringing it up now, which is, it's called bovine content. Being overly verbose in needless expositions. Which is the too much text. Bring it in.

(Joel Beasley at 00:22:01) I love it.

(Joel Beasley at 00:22:06) Oh, yeah. And then I also saw on Twitter today, somebody texted me this. But within like 24 hours of the regulation about the watermarks, there's like this massively popular open source project on GitHub to remove the watermarks. I was like, that is so brilliant. That is like, to me, I love, we're gonna put this, we're gonna spend millions of dollars, put this huge legislation together, pass it, take all this time, and then someone's like, open source project, one-click removal. Yeah. That makes me happy.

(Alan at 00:22:42) The community answers.

(Joel Beasley at 00:22:44) Yeah. Right. We're supposed to be talking about legacy modernization.

(Alan at 00:22:48) I think modernization is the topic a little bit.

(Joel Beasley at 00:22:51) Yeah, it was the colors. The color look. I got the blue and the purple. Alright. Legacy modernization. You have a ton of experience in this. What's hot? What's going on there right now?

(Alan at 00:23:05) Well, I think, I mean, if there was any area of the sort of tech space that was right for AI, it is the modernization of legacy code. It's taking what is an old stack and moving it into a much more modern stack that's more maintainable, etcetera. And, historically, that was a big problem to take on. It was a big undertaking to take on because, usually, that legacy platform was being used in production, and you just didn't wanna get it wrong. And we came up with, you know, common patterns like the strangler pattern and what have you to take off chunks of it and rewrite bits of it and what have you. And that works to a certain degree up until you've got yourself like the big database sitting in the middle that sort of controls everything, and that database may not be structured the best way it is either. So it's a real problem. And the time to do, the time that it takes to do all of that is, it takes a lot of effort for a product engineer and a systems engineer to dig in and figure out what all the business rules and some of the bizarre edge cases. And you think, why? Why are we doing that? Who was smoking what when we developed that particular piece? But clearly, that is still a very important piece because we've now sort of hinged a lot of workflow upon that. So AI can now help address that.

(Alan at 00:24:19) And sort of my argument in this, and it's played out in a number of projects now, is stop taking bits of the legacy system. The strangler pattern, I don't think, holds as much as it did in the new AI world. The argument now is, leave legacy where it is. Let it do what it's doing because, frankly, it's not your reference implementation. It's the one that everybody knows. The business is working on it, and there's no risk with it at the moment because it's doing what it's supposed to do. It may fall over every so often. It may have a little bit of idiosyncrasies, a little bit more handholding in certain areas, but the company has evolved around it to work with its personality in order to make it there. So, therefore, when one is looking to modernize that, replicate it in a new modern stack, and replicate it to the point where even the UX still looks the same because that sort of alleviates the change management initially. Because once we've got the underlying pinnings modernized, the UX becomes a lot less problematic to start to change out. So AI can really run at that old legacy code base, pull out all of the business cases, allow you to stand up a completely new architecture, allow you to question some of the business rules that is pulled out to determine, is that still relevant? Is that still something we wanna move forward with?

(Alan at 00:26:10) Because, ultimately, even with the old school way of modernizing legacy systems, you still had the data export/import problem at some point. You had the whole switchover from old to new problem. So if you accept that that's still gonna be a step in this process, then the AI solution helps you get to that modern stack much faster. It allows you to play a lot more testing with it because you can see what the old system did, and you can now see what the new system does. And you've literally got something where you have not touched anything. Whereas in the strangler pattern, sometimes you have to tweak something in the old system in order for it to, so you're also introducing noise into sometimes legacy code because you have to screw yourself into that sort of stuff. So from that perspective, AI has allowed us to play that sort of bigger game where, yeah, you're not having to craft out all of the code yourself. You can do this sort of in an iterative process to say, okay, just do a simple copy. Do a simple convert. Now let's iterate on the underlying architecture and make it better. But, ultimately, legacy platforms should not become that big gnarly, keep pushing it down the line project that we used to do. And particularly in the sort of day and age of buying and selling companies, we do want to see a lot less legacy systems being bought and sold. That's part of the value add from the private equity company is to be able to turn that little wagon around and effectively present the new owners with, okay, we've done this bit. Now you can go off and grow this company even further.

(Joel Beasley at 00:28:08) I wanna, I wanna make a correction. I went, ugh, at Zed. I was, I was thinking you were referencing the IBM Z mainframe systems, like the older mainframe stuff. Zed, I just Googled it because I was like, I don't think that was the right response. And it's like a, I use Cursor. It looks like a Cursor alternative.

(Alan at 00:28:30) Yes. Yes. Okay. Z-E-D.

(Joel Beasley at 00:28:34) Yeah. Zed code. Yes.

(Alan at 00:28:35) Zed code. Zed code.

(Joel Beasley at 00:28:38) Is it Zed code? Is that what it's called or just Zed?

(Alan at 00:28:41) I think it's Zed code, isn't it?

(Joel Beasley at 00:28:45) There's two. There's Zed and zed.dev. We're learning a lot. There's zed.ai.

(Alan at 00:28:54) The IDE is written in Rust.

(Joel Beasley at 00:28:57) Oh, wow. There's a lot of, this is, this is totally what engineers do, though.

(Alan at 00:29:04) You're going side tangents. We do. It's a feature, not a bug.

(Joel Beasley at 00:29:08) It is a, yes. It is.

(Joel Beasley at 00:29:12) Oh, man. Have you seen Blitzy? We're talking about products. Have you seen Blitzy? No. You gotta check Blitzy out. I interviewed them. So this company called me, and they're like, hey, we've got, they wanted to come on the show. And they were like, oh, we built this thing. This is like a year or two ago. And they were making all these claims that I was like, oh, this is ridiculous. And then I like looked up their founders. I was like, they're really smart guys. And so I told my producer, I was like, all right, let's have a meeting with them. And they made these like real bold claims about the AI agents, like, building these, basically rewriting these legacy systems and stuff. And so I met with them and their CEO was brilliant. You know, some Ivy League guys, like Brian and, super smart people. When they showed me it, and I was like, well, that's cool. And I said, well, give me one of your customers, you know, because you and I know about like how vaporware works and all this. So I was like, give me one of your customers, like a big customer. And they gave me one of, like, is one of the top four consultancies, you know, one everybody would know. I won't mention their name, but, and so we did an interview with them. And I have Brian come on and then the CTO of that big consultancy come on. And we talked about the good, the bad, the ugly. Dude, it is totally true. They've built the system and it's saving them like massive amounts of money. And to me, I'm like, when you find something like that, it's so rare that you just wanna tell everybody about it. But so check it out. It's called Blitzy.

(Alan at 00:30:48) Blitzy. Yeah. Okay.

(Joel Beasley at 00:30:50) Yeah. I don't know how they're describing themselves today, but I know that if I look them up real quick, they're called, they're describing themselves AI-powered autonomous software development. So nothing crazy, but they're working with a bunch of the enterprises. And that's all the stuff you described about how you go through that and you look at the legacy system and then you use that as the print to build the more modern, and then you pull the human in and ask them if things still, all of that was like you work at Blitzy and we were just doing a podcast. So when those connected.

(Alan at 00:31:23) They're happy with the free publicity. Now they get two bites of the Modern CTO podcast. But the reality is that AI has allowed us to play those games now, kinda like how cloud computing back in the days allowed us to play games with the server as if it was a function. Now AI is allowing us to go after this stuff. And it's all about just de-risking the overall project, which is I don't wanna necessarily jump into a half-assed new system that's legacy is powering some of it, and the new shiny toy is powering the other half of it. And now I've got a change management. And we've seen it so many times where you're using a product, and the UI suddenly changes into the cool looking new UI. And sometimes you go back to, I mean, Salesforce ran with that for years where it was still the old stuff where you would say, okay, yeah. We're in, we're in the old world now because this is the old, Web 1.0 UX that we're in. Because, again, it was so brittle. They just don't wanna touch it.

(Joel Beasley at 00:32:22) Yeah. I've noticed a lot of companies doing that. Like, I'll go to log in, and it'll be like, which version are you on? And, you know, honestly, as an engineer, I'm sure that was the right choice for them. Because if you don't have a team that's constantly updating and like pushing to production cons, if you don't have, it really boils down to what's the habits of the team that is building the original product and where do they wanna go, and how can you structure it? So I don't, I don't judge them necessarily. I just assume that they made the best decision for their current situation.

(Alan at 00:33:01) Yeah. For sure. But I think with the sort of the AI taking away a lot of the heavy lift of documentation, QA testing, all of that, all the ugly bits that developers, frankly, don't like to do. We all pretend we do it, but we don't, kinda like flossing our teeth. I'm very happy that AI has picked up that particular stuff. I mean, some of the enterprises that we're working with now, the dirty secret is they've never been better tested, you know, nowadays than what they were in the old days.

(Joel Beasley at 00:33:34) Days. What is leadership thinking about this? When we're talking about building and modernizing, how are they—

(Alan at 00:33:40) Gonna sword there. Yeah. Because they get hooked on the new accelerated speed and think that everything can be done at that speed. And one has to be very careful about, frankly, putting deadlines out and timelines out. I mean, we've all heard of the 10x developer. It's just nonsense. 10x what? The amount of code you generate and the amount of time—it's just a silly metric, but yet it's one that's used a lot at the C-level and the board level. So you have to really temper the expectations that, yes, AI will accelerate it and probably twice the speed of it, but let's not bank on that and see it more as a bonus as opposed to something that we're really gonna say, right, we're gonna have everything modernized for Christmas.

(Joel Beasley at 00:34:32) Yeah. But I want it done three months sooner, though, Alan.

(Alan at 00:34:35) Yeah. So inevitably, something's gonna come out of the woodwork that is just gonna—and it could be the fact that some of the AI tooling has shifted from what we started off before. I mean, you said you use Cursor. I use Cursor a lot. But last week, every day, menu items kept moving all over the shop, so I had to turn off daily updates on Cursor. And, you know, it was like, stop. So everything is moving in this space that is just—it's very hard to get your arms around it and plan long term.

(Joel Beasley at 00:35:15) Yeah. I definitely turn off those updates. I update when I have to, and then it's this massive change all at once. Yeah. That's what I'm doing. The executives, though. Okay. One of the things that I saw—a lot of the work I did in my career was rewriting existing systems. And one of the things that I always found was when we went to go do that, obviously pre-AI, the executives had to get involved. We would have to talk with them again about how the logic flows for the business, how decisions are made, and it would inevitably frustrate them. It's almost like they made the decision. They want to swipe it, put it under the rug, never think about it again, and then we're bringing up old stuff. Are you seeing that happen?

(Alan at 00:36:01) Yeah. Because I think where the legacy systems are now starting to bite is both in the retainment of—recruiting and effectively getting people to maintain the systems. And sometimes, some of these systems are not done in such a way that you really want AI to be running loose in there because there's just too much interdependencies and just illogical connections that made sense for the developer at the time. But in the cold light of day, AI will quickly correct it and say, no, you don't need it to be like this. And then, oh, we were relying on the fact that that word was misspelled over here to drive this piece of logic over there. And from that perspective, when—and particularly in my space where we are sort of putting lipstick on the pig in order to sell it to the next organization—it has now become a true discussion point because AI diligence space is now surfacing all these products and all this stuff and producing this huge report to say, yeah, and this was now discontinued in 2007. This was no longer supported. This is now—so there's nowhere to hide anymore. So we're having to face the fact that we kind of need to do this housekeeping.

(Joel Beasley at 00:37:24) What's the biggest mistake that you see CTOs making with modernizing?

(Alan at 00:37:30) Not underestimating the change management to the rest of the business, which is—while you can keep everything the same and what have you, there will be changes. But it's not bringing the rest of the company along with you. So it won't work well in a vacuum.

(Joel Beasley at 00:37:45) How do you do that well, though? Not how do you work well in a vacuum, though. How do you work well? How do you ignore people well? It involves turning your phones off.

(Alan at 00:37:55) Yeah. It just basically finding the—it's communicating to the company a language that they can understand with no buzzwords and articulating to them what it means to them. And what you have to be very careful about—and, you know, I sometimes think that the modern CTO has become also a therapist—is, developers, we've gone through that sort of change of, well, if AI's producing all the code, what value am I giving? What do I do? I actually quite like the coding part. And now you're telling me I don't need to do that part. And there's a crisis of their own sort of worth and their own value to which they're providing. So CTOs are already having that sort of, okay, come and lie on the couch. Let's talk about how we can use AI to be an accelerant for you as opposed to somebody that replaces you. And it's the same thing as one is communicating outside of the technical department—is position it in such a way that it doesn't feel threatening to anybody. It doesn't feel—because as soon as they hear AI, suddenly think, oh, will I still have a job at the other end of this? Will this process still be needed? And maybe not, but again, don't make people feel they have to fear change and fear what's coming.

(Joel Beasley at 00:39:28) I read a book.

(Alan at 00:39:30) That's a good start, Joel.

(Joel Beasley at 00:39:32) I did. What? I listened to a book. We already lost. So it talks a lot about—the subtitle of it is 10 Irrefutable Principles of Being Creative in the AI Era. And a lot of it was talking about, like, it's written for creatives and it's saying, alright, creatives, whether you're filmmakers or, you know, graphic designers or software engineers, you're creative. And here's how AI is intersecting with that. And it focuses purely on all of those questions you just mentioned. Like, are we the curators now? Is that valid? And it goes through several different—it's a pretty quick read or listen. It listen time was like four hours. I did it in a car ride. But it's called Artism Rules. I'll send it to you here. I hate when people recommend content. Watch this TV series. Read this book. Like, I don't—I am content out. Too much content. So when I do recommend something, it is easily the best book that I've read anywhere near this AI and creativity concept. I'll show it to you here. It's on Amazon.

(Alan at 00:40:56) Okay.

(Joel Beasley at 00:40:57) It's called Artism Rules.

(Alan at 00:40:59) Got it.

(Joel Beasley at 00:41:00) Yeah. And even if you don't read it, recommend it to somebody else. There you go. Because it's good. It's good. It is real good. So I gotta get him on for an interview is what I gotta do.

(Alan at 00:41:15) Good idea. Well, at least now you've read it. Listened to it.

(Joel Beasley at 00:41:18) Yeah. It's so funny. Sometimes I read these books and I'm like—or listen to them or whatever. I don't read. I listen. Do you know how much I read interacting with AI every day? Yeah. I'm reading pages and pages. Got all my reading done. We're off topic. Alan, I love hanging out with you. We made a podcast. How do you feel?

(Alan at 00:41:40) I thought it was wonderful. Thank you. I've taken a lot away from this. I think I've got more out of this than—

(Joel Beasley at 00:41:45) Untrue. You filled my day up. You made my day. Like, today, the podcast affects my day. It always does. And when I have a good podcast, the moment I heard your voice, within the first ten seconds of you speaking, I was like, oh, I like this person. I'm excited to be talking to them. This is gonna be a good day.

(Alan at 00:42:04) I'm quite proud of myself. I haven't sworn. So it's—I usually slip into the odd expletive. The Scots are known for that. We're passionate. So I'm taking that as a win.

(Joel Beasley at 00:42:16) That is a win. Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn, or send me an email [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.