[00:00:00] Welcome back to another episode of
[00:00:01] TechOnHinch, where tech gets human
[00:00:02] powered by Code District. I’m your host
[00:00:04] Robia Jabed and my guest today is James
[00:00:06] Seddon, a technology executive and
[00:00:07] founder with more than 25 years of
[00:00:09] experience building and scaling software
[00:00:11] from IBM, Motorola, and McDonald’s to
[00:00:14] Deloitte and now his own company, Soft
[00:00:16] Exit. James, welcome to TechOnHinch.
[00:00:18] It’s great to have you on the podcast.
[00:00:20] >> Thank you, Robia. Nice to be here.
[00:00:22] >> We see that most companies just use AI
[00:00:24] to build apps. You build the tool that
[00:00:26] does the building, where someone
[00:00:28] describes what they need in plain
[00:00:29] English and it creates the app for you.
[00:00:32] What’s the real difference between those
[00:00:33] two?
[00:00:34] >> The real difference is everyone is
[00:00:36] using, as you point out correctly,
[00:00:37] they’re they’re building apps, but it’s
[00:00:39] generating code, right? And it’s also
[00:00:41] generating bugs and security issues and
[00:00:44] all sorts of things that come with it,
[00:00:45] right? And it has feature drift, where
[00:00:48] you lose features, right? You have to
[00:00:49] constantly do regression test every time
[00:00:51] you ask for a build from AI, you’re
[00:00:53] getting a whole new application,
[00:00:55] regardless of somebody wants to
[00:00:56] recognize that or not. So, you have to
[00:00:58] really do full regression test. That
[00:01:00] seemed laborious to me and basically, I
[00:01:04] created a system where we have a
[00:01:05] platform that already exists and it, you
[00:01:07] know, an individual can go in and
[00:01:09] customize it and make it to their needs,
[00:01:11] but if AI were to be the assistant and
[00:01:14] be able to configure all that, it
[00:01:15] configures it on a stable platform that
[00:01:17] has been thoroughly tested, that they
[00:01:19] don’t have to worry about bugs, that the
[00:01:21] worst thing that could possibly happen
[00:01:23] is something got misconfigured and they
[00:01:25] can go in and reconfigure it without
[00:01:28] having to generate new code and without
[00:01:30] creating more exposure to more defects
[00:01:32] and security issues. So, that’s the
[00:01:34] approach we took. It’s not generating
[00:01:36] application, it’s generating an
[00:01:38] application on a platform and that makes
[00:01:40] it way more predictable and stable.
[00:01:42] >> So, James, it’s been, you know, around
[00:01:43] 20 years that you’ve been in the, you
[00:01:45] know, software industry, from big
[00:01:47] companies like IBM and McDonald’s to
[00:01:49] your own startups. Is AI actually
[00:01:51] changing how you build software or is it
[00:01:54] just making the old way faster? What do
[00:01:56] you think?
[00:01:57] >> Oh, 100% change. It’s completely
[00:01:59] different. The previous things that I
[00:02:01] did on my own, I literally wrote the
[00:02:03] code. I’ve been writing code since the
[00:02:04] beginning of my career. I’m really good
[00:02:06] at it. I mean, I would hope after doing
[00:02:08] it for so long that I am good at it.
[00:02:09] Now, I don’t write code anymore.
[00:02:11] Honestly, I’ve become more of the
[00:02:13] director in this orchestra here and it’s
[00:02:16] very much like when I’m working for a
[00:02:17] company when I’m not writing code. I’m
[00:02:19] giving direction to engineers. I’m
[00:02:21] telling them explicitly how I want it
[00:02:23] designed, explicitly how I want the data
[00:02:25] to be set up, the communications,
[00:02:27] everything. I actually kind of upset
[00:02:29] architects because I involve myself so
[00:02:31] deeply in the design of the system that
[00:02:33] I find that I’m performing that exact
[00:02:35] same role with AI. But, AI becomes the
[00:02:38] coder. That’s the real difference.
[00:02:40] >> The pace is so fast that every day there
[00:02:42] is some new feature, a new update, and
[00:02:45] you just got to keep up, you know, with
[00:02:47] all of that. Otherwise, you’re probably
[00:02:49] outdated or you’re not going to make it.
[00:02:51] >> It depends on where you’re going, right?
[00:02:52] If you want to work for a corporation
[00:02:54] and you want to impress them with your
[00:02:56] AI knowledge, you got to know all the
[00:02:57] latest, right? You got to know Cloud,
[00:02:59] you got to know Chat GPT, you got to
[00:03:00] know all of that stuff. But, it’s a
[00:03:02] different perspective when you’re
[00:03:03] building it for yourself. You really
[00:03:05] don’t need to learn all that because
[00:03:07] technically, they’re all the same.
[00:03:09] Agents and all that, they’re all a
[00:03:10] disguise of the LLM. And my theory is,
[00:03:14] now I’m sure people will tell me I’m
[00:03:15] wrong, but I really believe I’m right
[00:03:17] though without any facts, is I think
[00:03:19] that Cloud code is what they have is
[00:03:22] they have the same LLM engine, right? I
[00:03:24] mean, they didn’t just invent they left
[00:03:27] Chat GPT overnight. So, they basically
[00:03:29] utilize the same thing and then they
[00:03:31] prompt it and, you know, you have the
[00:03:33] instructions, right? You have the
[00:03:35] skills. I think those are all just
[00:03:36] prompts that you don’t have to pay for
[00:03:38] resubmitting over and over and they put
[00:03:40] a wrapper around the LLM and they’re
[00:03:42] calling it, “Hey, these are agents.
[00:03:43] These are new.” And agents are just
[00:03:44] iterative prompts. So, it’s really the
[00:03:46] same thing. And if you approach it that
[00:03:49] way, you might not have to pay for all
[00:03:51] those fancy bells and whistles. You can
[00:03:52] do it yourself on a $20 account.
[00:03:54] >> In 2026, what do you think that AI, you
[00:03:57] know, genuinely can build on its own
[00:04:00] from start to finish? And where does it
[00:04:03] still lag or fall apart?
[00:04:05] >> So, I think the answer to your first
[00:04:06] question will answer your second. What
[00:04:08] What can it build all on its own?
[00:04:09] Absolutely nothing.
[00:04:11] I really I think it is a great coder. I
[00:04:14] think it completely lacks long-term
[00:04:18] planning, long-term designing. And that
[00:04:20] I think if you approach something in
[00:04:22] small projects, like you say, “Build me
[00:04:24] this function.” It is fantastic at it.
[00:04:27] If you say, “Build me Salesforce.”
[00:04:29] Forget it. I mean, it’s just not going
[00:04:31] to happen. So, it’s not going to happen
[00:04:33] to the level that you won’t iterate and
[00:04:34] iterate and iterate and iterate. So, you
[00:04:36] probably would have been better off
[00:04:38] designing it a piece at a time rather
[00:04:40] than asking it to build the world, then
[00:04:42] rebuild the world, then rebuild the
[00:04:43] world. Do you know what I mean? That’s
[00:04:45] my impression of it.
[00:04:46] >> So, you know, if we get a bit practical,
[00:04:47] James, where are you actually using AI
[00:04:50] agents at the moment? What work are they
[00:04:53] doing for you? And what problem are they
[00:04:55] really solving?
[00:04:56] >> I don’t use agents at all. I really
[00:04:58] don’t. Right now, I’m not trying to
[00:04:59] automate sales leads and all that stuff.
[00:05:01] So, maybe I’ll come back and look at
[00:05:02] that. I I don’t think I would build
[00:05:05] those out. I think I would look towards
[00:05:07] the third parties that have already
[00:05:09] figured all that out and just kind of
[00:05:11] connect them all together. But for
[00:05:12] building an application, I don’t use
[00:05:14] agents. I literally am down to the bare
[00:05:17] basics of where I use ChatGPT and I have
[00:05:20] a conversation with it and I build it a
[00:05:21] function at a time. And it saves me all
[00:05:24] the AI tokens and it gives me better
[00:05:26] quality code, more control over the
[00:05:28] design. And this is just my method. I’m
[00:05:30] sure everyone out there is like, “Well,
[00:05:32] he doesn’t know what he’s doing.” Okay,
[00:05:33] that’s fine, but I’ve built something
[00:05:35] pretty incredible by doing that.
[00:05:36] >> you talk about, you know, what have you
[00:05:38] tried building with it? Or is it
[00:05:40] something that you sort of plan to do?
[00:05:42] >> Well, actually, I’m probably a month
[00:05:44] away from releasing Softeksit Studio,
[00:05:47] which essentially is a platform an
[00:05:50] enterprise platform for companies. It’s
[00:05:52] got a dynamic database. It’s got
[00:05:54] dashboards. It’s got It’s everything’s
[00:05:56] very familiar, right? But we have low
[00:05:58] overhead, so we’re going to come out and
[00:05:59] offer this at, you know, $14 a month per
[00:06:02] user. So we’re going to really catch it
[00:06:04] on cost. But, you know, I don’t want to
[00:06:06] go into a sales pitch here, but it’s way
[00:06:08] easier to use. It’s beautiful looking.
[00:06:10] It’s It’s got incredible features that
[00:06:13] require almost no real technical
[00:06:15] knowledge from people. Like real
[00:06:17] relational database. All you have to do
[00:06:18] is say, here’s the parent and child and
[00:06:20] it’s done. Everything else adjusts all
[00:06:21] the forms and everything. So all this
[00:06:23] was built with AI and it was built with
[00:06:26] AI in mind. So at the end that the AI
[00:06:29] architect that we have built into the
[00:06:31] platform will be able to configure it
[00:06:33] all because it they speak the same
[00:06:35] language. There’s really no translation.
[00:06:37] It’s pretty amazing.
[00:06:38] >> There was a study by MIT
[00:06:40] of experienced developers using AI. They
[00:06:43] thought it made them faster, but it
[00:06:45] actually made them slower and they
[00:06:46] didn’t really notice. When your own team
[00:06:48] says AI is helping, how do you check
[00:06:51] whether it really is helping them?
[00:06:52] >> Well, I think the study is based upon
[00:06:55] they reviewed all the check-ins for
[00:06:57] GitHub and and things like that. And I
[00:06:59] believe, you know, we’re still in the
[00:07:01] infancy of this. I think they just
[00:07:03] turned developers loose and said, “Hey,
[00:07:04] go at it. Here’s AI. Make yourself more
[00:07:06] productive.” with no direction and they
[00:07:08] all went in that direction and just
[00:07:10] created a ton of iterations which
[00:07:12] actually slowed them down. So since they
[00:07:14] were encoding, it felt it was faster,
[00:07:16] but the data showed they had way more
[00:07:18] iterations and in the end they were
[00:07:20] about 20% less productive than they
[00:07:22] thought they were more They thought they
[00:07:23] were 20% more productive. I can clearly
[00:07:26] see how that would happen. I think you
[00:07:28] need new type of roles like AI
[00:07:30] architects to be able to come in and
[00:07:32] give them direction on how to use the
[00:07:34] tool as opposed to just giving them an
[00:07:36] order, “Make yourself more efficient.”
[00:07:38] because I think leadership doesn’t fully
[00:07:40] understand this yet and all they want to
[00:07:42] do is communicate to their their their
[00:07:44] C-level leadership that hey, we’re doing
[00:07:46] AI, right? Cuz everyone’s got to check
[00:07:48] that box, but no one is saying what the
[00:07:50] directive is of the results of that box
[00:07:53] being checked.
[00:07:53] >> There’s this old rule in software. The
[00:07:56] first 90% of the work takes 90% of the
[00:07:58] time, and the last 10% takes another
[00:08:01] 90%. So, what do you think? AI has made
[00:08:04] the first 90% almost free. That’s what
[00:08:06] they’re saying out there. Has it done
[00:08:08] anything for the hard part at the end to
[00:08:10] make it this clear?
[00:08:12] >> Not at all. And there’s two reasons.
[00:08:14] There’s technical, and then there’s
[00:08:16] reality, right? AI could generate a
[00:08:18] perfect application. Because I know
[00:08:20] this, without AI, I have generated many
[00:08:22] perfect application. And you bring it to
[00:08:24] leadership, and you you’re going to get,
[00:08:26] you know what? I don’t like this font. I
[00:08:28] don’t like where you put this menu. But
[00:08:29] it functions perfectly, but they don’t
[00:08:31] like it. And getting through that
[00:08:33] process is an extra 35%. I guarantee on
[00:08:36] every project is this constant iteration
[00:08:39] of opinions, right? And you’re going to
[00:08:41] AI is not going to eliminate that.
[00:08:43] Leadership is going to have their say,
[00:08:44] and they’re going to have their
[00:08:45] opinions, and then you have your
[00:08:46] customers who are not going to like the
[00:08:47] way it was done, right? So, as you go
[00:08:49] through your focus groups, you’re going
[00:08:50] to go through more iterations of that.
[00:08:52] You’re still you still have the same
[00:08:54] problem of your BAs are going to give
[00:08:56] you your input, you’re going to design
[00:08:58] it, and it’s not always going to work
[00:08:59] out the way it was asked, and you’re
[00:09:01] going to still have that stabilization
[00:09:03] as well. Stabilization with bugs. You’re
[00:09:05] going to still going to have bugs, no
[00:09:06] matter what. So, none of that goes away.
[00:09:08] So, that last 90% is is usually going to
[00:09:11] take the longest. The saying I’m used to
[00:09:13] saying is, well, we’re done, but we’re
[00:09:14] not finished. That’s that’s been my
[00:09:16] favorite one.
[00:09:17] >> [laughter]
[00:09:17] >> Yeah, makes sense. James at Deloitte who
[00:09:20] did federal work where every requirement
[00:09:22] was written down before anyone built
[00:09:23] anything. So, the industry spent years
[00:09:25] trying to get rid of that particular
[00:09:27] step. But AI can only build what you can
[00:09:30] clearly describe to it. Are we about to
[00:09:32] need that particular skill back? We kind
[00:09:34] of do all that.
[00:09:34] >> Absolutely.
[00:09:36] That’s That was a really insightful
[00:09:38] point you brought up. I’m telling you, I
[00:09:40] have been talking to some of my peers
[00:09:42] and I said, you know, we have done
[00:09:44] ourselves a disservice with Agile
[00:09:45] programming. We’ve done away with all
[00:09:47] the documentation and all that. It was
[00:09:49] great. And then we traded all that for
[00:09:50] micromanaging of developers, right? And
[00:09:52] everybody thought this was a great idea.
[00:09:54] I came from also Motorola and that was a
[00:09:57] CMMI level five organization in Motorola
[00:09:59] Solutions. Incredibly well run. And they
[00:10:01] had very defined processes for software
[00:10:03] development. And I remember the phrase
[00:10:05] where they would say they would use
[00:10:06] Rational Rose Suite. I don’t know if
[00:10:08] you’re familiar with that, but it is IBM
[00:10:10] should really dust that off and
[00:10:12] repurpose it for AI now because that
[00:10:14] tool would be amazing. So, the concept
[00:10:16] was that they would design the
[00:10:18] application so far that the phrase was
[00:10:21] the code would write itself. And you
[00:10:23] would have all the requirements, they
[00:10:24] would define it right down to the
[00:10:26] function. So, the developers had very
[00:10:27] little creativity. They would be given
[00:10:29] They would be assigned a function with
[00:10:30] the inputs and the outputs and they
[00:10:32] would have to write to that function.
[00:10:33] So, all this was designed in Rational
[00:10:35] Suite and it was very kind of process
[00:10:37] heavy. I think if we did that, well,
[00:10:40] guess what? The phrase, well, the code
[00:10:42] will write itself, we have achieved the
[00:10:44] code will write itself actually. So,
[00:10:47] having purely defined requirements fed
[00:10:50] into an AI engine would be absolutely
[00:10:53] amazing. And I really think people
[00:10:55] should go back and look at the Rational
[00:10:56] Rose Suite and reconsider building those
[00:10:59] type of applications because in my
[00:11:01] opinion, coding is virtually gone.
[00:11:04] >> that’s some of the people who were
[00:11:05] sitting at the top, you know, that’s
[00:11:07] what they were saying that, you know,
[00:11:08] who cares about the code quality anymore
[00:11:10] as long as it’s giving you what you
[00:11:12] want.
[00:11:12] >> I’ll throw one thing in there though. I
[00:11:14] agree because, you know, when I went
[00:11:16] through coding to management, you know,
[00:11:17] I had a I struggled, right? cuz not you
[00:11:19] promote your best code and your best
[00:11:21] coder is has high standards and is
[00:11:24] expecting everyone else to code and at
[00:11:26] that level and I really struggled with
[00:11:27] the fact that developers were not as
[00:11:29] good as me at the time, right? And And
[00:11:31] this is just levels. This is not This is
[00:11:33] not pride. This is just a senior
[00:11:35] engineer looking at entry-level. So, the
[00:11:37] thing is is the only way you can
[00:11:39] compensate with that is process and
[00:11:42] defining your data and defining the
[00:11:44] structure and have them code to it. So,
[00:11:46] if you let them design a bad data
[00:11:48] structure and a bad architectural
[00:11:50] design, it doesn’t matter how good your
[00:11:51] code is. You can’t dig yourself out of
[00:11:53] that. But, if you define your data
[00:11:55] really well and efficient and clean
[00:11:57] tables, clean relationships between data
[00:12:00] and clean design of the architecture, if
[00:12:02] the code works and it passes, it’s okay.
[00:12:05] If the performance isn’t there, then you
[00:12:06] have a performance problem and you deal
[00:12:08] with it. But, if the code works, it’s
[00:12:10] fine cuz you can always clean up the
[00:12:12] code, but you can’t clean up a design
[00:12:14] easily cuz it has a domino effect. So, I
[00:12:16] really think that people need to really
[00:12:18] start focusing on design now and get And
[00:12:20] don’t worry about what the code is being
[00:12:22] generated, right? And that I’m sure
[00:12:24] programmers are going, “Oh my god, I
[00:12:25] totally disagree.” But, the truth is is
[00:12:27] when you branch into management, you
[00:12:29] can’t control the quality of code,
[00:12:31] especially if you’re on a large team. If
[00:12:33] you have a thousand developers, are you
[00:12:34] going to code review a thousand
[00:12:36] developers’ code? You can’t. So, you
[00:12:37] have to settle on a few things. You have
[00:12:39] to settle on process and you have to
[00:12:41] settle on design. And then the rest can
[00:12:43] fall into place. Does that make sense?
[00:12:44] >> Yeah. Yeah, it does. So, you know,
[00:12:46] James, when you ran a credit card
[00:12:48] processing, the kind that gets audited
[00:12:51] every year where a mistake is really
[00:12:53] expensive. On papers, it’s pretty
[00:12:56] strict. There are clear rules which
[00:12:57] sound perfect for AI to build in. But,
[00:13:00] is that actually the best place to trust
[00:13:02] AI or or the worst?
[00:13:03] >> Well, I mean, AI is just like another
[00:13:05] coder, right? You know, you got a big
[00:13:06] team. You Again, like I said, you’re not
[00:13:08] going to be able to review everyone’s
[00:13:09] code. So, what you need to rely on is a
[00:13:11] good QA department, right? And they’re
[00:13:14] the ones that are going to find all you
[00:13:17] know, the bad code, the the security
[00:13:19] holes and all that. And then So, but now
[00:13:21] that’s becoming more automated and I
[00:13:23] haven’t looked at that, but that sounds
[00:13:25] very promising because if you could
[00:13:26] build an entire application from AI and
[00:13:30] then you can do a full regression test
[00:13:32] with AI, now you really got something
[00:13:34] because now you got both ends. You can
[00:13:36] you can trust it better and I believe
[00:13:37] eventually AI can test way better than a
[00:13:39] person because testing is a difficult
[00:13:42] job because you’re doing the same thing
[00:13:43] over and over and over again and then
[00:13:45] like at McDonald’s, they had you know,
[00:13:48] very strict security requirements which
[00:13:50] we met but we had to hire big QA teams
[00:13:53] and after a while we would cycle them
[00:13:55] because they would not see the defects
[00:13:56] in front of them because it’s human
[00:13:58] nature. You keep doing the same thing
[00:13:59] over and over again. You’re just going
[00:14:00] to you’re going to not have the same
[00:14:02] attention to detail and then so we would
[00:14:04] cycle them where AI won’t have that
[00:14:05] problem. It won’t get fatigued. It won’t
[00:14:07] get It won’t lose sensitivity over what
[00:14:09] it’s seen over and over but I haven’t
[00:14:11] seen a lot of AI testing suites yet.
[00:14:13] >> Well, I got to ask now that you
[00:14:14] mentioned, did you get free Happy Meals
[00:14:16] when you were working at McDonald’s?
[00:14:18] >> I did not and you know, I’ll tell you
[00:14:21] they didn’t even give us a discount on
[00:14:22] the food because they said the profit
[00:14:24] margins were so thin. I understand that.
[00:14:26] >> That’s it.
[00:14:27] >> They did feed us well when they had a
[00:14:28] meeting though. So
[00:14:29] >> Oh. All right. So if a business person
[00:14:31] can just describe an app and have AI
[00:14:34] built it, what happens to the developer
[00:14:36] and to the whole custom software
[00:14:38] business people built their careers on?
[00:14:40] >> I think software developers need to
[00:14:41] figure out where they’re going. I really
[00:14:43] do. The more I work with it, the more I
[00:14:44] realize we’re really not going to be
[00:14:46] hiring software developers and that’s a
[00:14:48] huge impact on the industry. Me as a
[00:14:50] software developer, it breaks my heart.
[00:14:52] I think the way you asked the question
[00:14:53] is business people asking for apps to be
[00:14:56] developed. I think that is a formula for
[00:14:58] absolute failure. There’s still other
[00:15:00] people involved in building software.
[00:15:02] There’s the the architect, right? There
[00:15:04] is the the subject matter experts. There
[00:15:06] is the technical leaders. I don’t think
[00:15:09] you can do away with any of those
[00:15:10] people. I think they’re greatly needed
[00:15:13] to guide AI into building a properly
[00:15:16] designed system and I think the only
[00:15:18] real role at risk here is obviously
[00:15:20] entry level and obviously mid-level and
[00:15:23] a lot of the worker bees that if I was
[00:15:25] going to hire a development team, I
[00:15:27] probably would hire only one person who
[00:15:30] is a senior person in each area of
[00:15:33] discipline. So, like back-end,
[00:15:34] front-end, mobile, I would hire three
[00:15:36] developers and I’d say, “Here’s how
[00:15:38] you’re going to use AI.” And I would
[00:15:39] instruct it on how to approach it like
[00:15:41] the way I’m using it. I’d have an
[00:15:42] architect involved. I’d have maybe a DBA
[00:15:45] cuz like I said, the data structures are
[00:15:47] far more important than the code now.
[00:15:49] Because the data structures dictate your
[00:15:51] entire design. So, the DBA has now
[00:15:53] become a superstar. The architect is a
[00:15:55] superstar and then some of your, you
[00:15:57] know, lead technical people that could
[00:15:59] actually pick technologies. And I think
[00:16:01] that’s really important now. Like for
[00:16:03] example, I would say what I do is I have
[00:16:05] it use Java for everything cuz Java is
[00:16:08] so fast and Java is no longer an issue
[00:16:11] where I need to find a top-end engineer
[00:16:13] really understands it because the code
[00:16:15] is going to be generated. Rather than
[00:16:17] picking something that’s easier for
[00:16:18] humans to use now, I pick something
[00:16:20] that’s faster and more efficient running
[00:16:23] because I get an incredibly performing
[00:16:25] application.
[00:16:25] >> So, you know, James, there’s also a
[00:16:27] growing argument around that regular
[00:16:29] SaaS is dead or dying. So, how does that
[00:16:33] hold up to you and how do you think
[00:16:35] about that?
[00:16:36] >> So, I’ve been, you know, as I’m building
[00:16:38] the SaaS, I have been watching a lot of
[00:16:40] YouTube and stuff and they’re all all
[00:16:41] the ones that say SaaS is dead. So, I’m
[00:16:43] like, I better watch that because I’m
[00:16:45] building that. I don’t think it’s
[00:16:46] completely dead. I think you will see a
[00:16:48] reduction in some companies like, you
[00:16:50] know, Salesforce is astronomically
[00:16:52] expensive and it doesn’t make any sense.
[00:16:54] It’s just a database and they charge an
[00:16:56] incredible amount for storage. They
[00:16:58] charge an incredible amount for just
[00:17:00] storing records. And you know, anyone
[00:17:02] technical has got to be scratching their
[00:17:04] head, “How do they get away with those
[00:17:05] prices?” But so I think you’re going to
[00:17:07] see companies start to look at that as a
[00:17:10] way to cut costs in building their own
[00:17:12] items. But there’s still, you know, a
[00:17:15] high bar of entry for a lot of
[00:17:17] underserved companies, right? So, you
[00:17:19] say to a company, take I came from the
[00:17:21] trucking industry, a CDL school, the
[00:17:23] largest one in the country. And the I
[00:17:26] started to work with people in this area
[00:17:28] of work. And if you go to them and say,
[00:17:30] you know what, you could build your own
[00:17:31] system and AI, right? They’re going to
[00:17:33] be like, well, how do I host that,
[00:17:34] right? That seems kind of hard. They
[00:17:36] don’t know how to do that. There’s a
[00:17:37] whole business sector, even though it’s
[00:17:39] available to them, they’re not going to
[00:17:41] be able to do it. Because even though
[00:17:43] you can build the app, you got to host
[00:17:44] it. to deal with the security. You got
[00:17:46] to do all that. Some people are not
[00:17:47] willing to take that on. So, maybe I
[00:17:49] just answered my question why people go
[00:17:51] to Salesforce is because then it’s all
[00:17:53] on the system and they don’t have to
[00:17:54] manage it and they just pay a lot of
[00:17:56] money for, you know, a simple database
[00:17:58] that if they had the knowledge, they
[00:18:00] could have hosted it themselves, but
[00:18:01] they don’t. So, I think AI is the same
[00:18:03] thing. Even though for for SaaS, right?
[00:18:06] Even though they can build it, the real
[00:18:08] question is is will they, can they? And
[00:18:10] then the other question is should they,
[00:18:12] right? Even though you can, should you?
[00:18:14] I could build my own car, right? But,
[00:18:16] you know, why would I do that? I could
[00:18:17] go buy one.
[00:18:18] >> Yeah, so also there’s a lot of debate
[00:18:19] around accountability these days, you
[00:18:21] know, when we talk about AI. So, when AI
[00:18:24] builds a business application on its own
[00:18:26] and it gets something, you know, wrong,
[00:18:28] you know, once it’s live, who’s really
[00:18:30] accountable there?
[00:18:31] >> How much of that is it for the
[00:18:32] developers? If a developer screws up,
[00:18:34] who’s really accountable? I It’s It’s
[00:18:36] the company. I mean, there’s That
[00:18:38] doesn’t change. It’s not like I’m doing
[00:18:40] a self-driving car and a policeman pulls
[00:18:42] me over and says, you were speeding.
[00:18:43] Well, I wasn’t driving.
[00:18:45] So, it’s not that, you know, cut and
[00:18:47] dry. It’s you’re still responsible. You
[00:18:50] know, you designed it. You You built it.
[00:18:52] You hosted it. You said it was secure,
[00:18:54] right? So, the company is responsible,
[00:18:56] just like they were before.
[00:18:57] >> That’s a really interesting point that
[00:18:58] you made there. So, what if we just
[00:19:00] forget about the hype for a second,
[00:19:02] which AI tools do you use personally,
[00:19:05] you know, for for your everyday work and
[00:19:07] couldn’t work without at this particular
[00:19:09] moment? And which ones should every CTO
[00:19:12] or CIO should have their hands on?
[00:19:14] >> I worked exclusively with ChatGPT and
[00:19:16] I’ll tell you I’m there may be other
[00:19:18] advancements in areas, but you know, I
[00:19:20] keep reading that they’re all going to
[00:19:21] be the same. It’s all going to be
[00:19:23] commoditized and it doesn’t matter
[00:19:25] really which one you go because they all
[00:19:26] seem to copy each other. So rather than
[00:19:28] spending my time or wasting my time
[00:19:31] looking at every single thing cuz you’re
[00:19:33] right, it’s very difficult to keep up
[00:19:35] with and there’s a lot of shiny objects
[00:19:36] to chase after that I focused on getting
[00:19:39] my stuff done. And so I just stuck with
[00:19:42] ChatGPT. I didn’t even I don’t even use
[00:19:44] their agents. I I literally just have
[00:19:46] the conversation. One thing I like about
[00:19:48] ChatGPT is it really knows me. It’s a
[00:19:50] little scary and sometimes and it starts
[00:19:53] to understand my style and it becomes
[00:19:54] like basically I would call ChatGPT my
[00:19:57] co-founder. I really would. I don’t need
[00:19:59] a co-founder. I have this tool and it
[00:20:01] literally does everything that I want it
[00:20:03] to do and has knowledge that I don’t
[00:20:05] possess and that’s exactly what you want
[00:20:07] on a co-founder is somebody who
[00:20:09] understands. The one thing I suggest if
[00:20:10] if you go down that path is ask ChatGPT
[00:20:13] to be brutally honest because it’s it
[00:20:15] likes to please and just tell it to
[00:20:17] attack your ideas, attack your designs
[00:20:19] and it will make you stronger, guarantee
[00:20:21] you, and your apps as well.
[00:20:23] >> It was very interesting for me James to
[00:20:25] you know, you telling that you are still
[00:20:27] using ChatGPT majorly, but you know, in
[00:20:29] the world of cloud code, you know, it’s
[00:20:32] still very rare to find people who are
[00:20:33] just sticking to ChatGPT or probably
[00:20:35] loyal to ChatGPT.
[00:20:36] >> Do you Do you know what scares me about
[00:20:37] cloud code is my wife will tell you this
[00:20:39] this is actually kind of a funny story.
[00:20:41] When I was originally working with
[00:20:42] ChatGPT, I started to trust it and I
[00:20:45] really wasn’t doing a lot of backups and
[00:20:47] I started coding and then I discovered
[00:20:49] feature drift and it just killed like a
[00:20:51] feature I spent a month working on and
[00:20:53] then I lost my temper and I was kicking
[00:20:55] cardboard box. I was screaming. I go
[00:20:58] language. My wife says, “What’s going
[00:21:00] on?” I go, “This thing screwed me.” I
[00:21:02] was so mad at it that I couldn’t believe
[00:21:04] it. Why I I use like cloud code. It’s
[00:21:06] cuz of feature drift. That feature
[00:21:08] scared me so much that I lost code. The
[00:21:11] thought of turning my entire code base
[00:21:13] over to cloud and it can edit anything
[00:21:15] it wants scares the crap out of me. I
[00:21:16] mean, I’m just like God knows what it’s
[00:21:18] doing. At least when I’m cut and pasting
[00:21:20] code in out of a chat, I know what it’s
[00:21:23] doing. I know the areas I need to test.
[00:21:25] I know the areas I need to focus on that
[00:21:27] I won’t take that step. Now, I would
[00:21:29] imagine I don’t fully understand cloud
[00:21:31] because I have limited myself to chat
[00:21:34] GPT so I can stay focused. But the
[00:21:36] thought though of in turning over an
[00:21:38] entire source code base to an AI and
[00:21:40] have it edit anything it wants is just
[00:21:43] very unsettling to me that I I don’t
[00:21:45] think I don’t think I have that kind of
[00:21:46] trust with feature drift cuz feature
[00:21:48] drift is far more painful for me than a
[00:21:50] bug. Bug I can fix. Feature drift is
[00:21:52] where a feature disappears that was
[00:21:54] coded and I don’t know it disappeared
[00:21:56] then I go to demo it and it’s gone.
[00:21:58] That’s bad.
[00:21:59] >> you deal with AI hallucinations when
[00:22:01] you’re working?
[00:22:02] >> I don’t know. I don’t really see
[00:22:04] hallucinations in code. I think
[00:22:05] hallucinations are more like if you’re a
[00:22:07] lawyer and you ask it to, you know, cite
[00:22:09] a case and it makes up a case, right?
[00:22:11] So, the code it either works or it
[00:22:13] doesn’t. It’s just like working with a
[00:22:14] developer. He’s either good or he’s not.
[00:22:16] You’re going to know.
[00:22:17] >> You know, James, five years out, do
[00:22:19] developers become the people who guide
[00:22:20] and edit AI or does part of building
[00:22:23] real software stay human? What are you
[00:22:26] betting on?
[00:22:26] >> I you know, the industry is creating a
[00:22:29] quite a paradox that no one’s going to
[00:22:30] go into coding. So, there no one’s going
[00:22:32] to grow their coders, right? So, there
[00:22:34] there’s no more entry level and the mid
[00:22:36] levels will eventually become seniors
[00:22:38] and these and the senior developers will
[00:22:39] basically become your technical
[00:22:42] architects at a lower level. You’ll have
[00:22:44] your senior architect and then you’ll
[00:22:45] have your lower level architects cuz
[00:22:47] they understand code and they understand
[00:22:49] a little bit how to design. So, I don’t
[00:22:50] know where we’re going because if once
[00:22:53] you run out of those people in about 10
[00:22:55] years or so, where are you going to grow
[00:22:56] people from, right? I think that’s a
[00:22:58] systemic problem. So, I think right now
[00:23:00] everyone’s experiencing a heyday that it
[00:23:01] can pick and choose who’s going to be
[00:23:03] their lead developers and designers.
[00:23:05] They can reduce their teams, but they’re
[00:23:07] actually kind of messing up the industry
[00:23:09] in the long run. So, I don’t know how
[00:23:11] that plays out because then you’re going
[00:23:13] to start to fall into what we were
[00:23:14] discussing earlier where the business
[00:23:16] people are asking for an application and
[00:23:18] they have no idea about tech and the
[00:23:19] thing creates some god awful pretty
[00:23:22] pretty looking monster underneath,
[00:23:24] right? And you know, there’s no
[00:23:25] foundation to the house and I think
[00:23:27] you’re eventually going to hit that
[00:23:29] path. So, I don’t know. I don’t know
[00:23:30] what happens there. So, long term, I
[00:23:32] don’t know. Short term, it’s crazy.
[00:23:34] Things are happening. It’s wild. But at
[00:23:36] some point, we’re going to plateau out
[00:23:38] on the knowledge and either AI has to
[00:23:40] catch up and really figure that out or
[00:23:42] going to be in a world of hurt. I think
[00:23:43] that people will start to go more the
[00:23:45] way I did. They’ll start to use AI on
[00:23:47] platforms that they can trust instead of
[00:23:50] having it generate pure code. I really
[00:23:51] think that’s the way to go. It’s going
[00:23:53] to be building blocks.
[00:23:54] >> So, after all these years, James, what’s
[00:23:56] the lesson about building software you
[00:23:57] had to unlearn the hard way?
[00:23:59] >> Unlearning is Yeah, I I spoke of it
[00:24:01] before is when I got into leadership is,
[00:24:04] you know, I I can’t expect people to be
[00:24:06] great code designers and have this vast
[00:24:09] architectural knowledge, right? I’m a
[00:24:11] full stack developer. It gives me
[00:24:13] incredible advantage. I’ve written
[00:24:15] native iPhone apps, Android apps, back
[00:24:17] end, front end. And my career started in
[00:24:19] database. I’ve been a coder in dozens of
[00:24:22] languages. I have a tremendous amount of
[00:24:24] experience that I can leverage. You
[00:24:26] know, expecting that from developers and
[00:24:28] even sometimes architects is something I
[00:24:30] had to step back and say, “Okay, not
[00:24:32] everyone’s going to have these skills,
[00:24:34] right?” So, and then the other thing I
[00:24:35] learned was I’m never going to find
[00:24:37] someone or Well, I mean, they exist, but
[00:24:39] they’re hard to find the unicorns who
[00:24:41] are full stack developers and have all
[00:24:42] this experience. So, we’re still going
[00:24:44] to have to have a team, right? It’s I
[00:24:46] don’t see I keep hearing people talk
[00:24:48] about the one company millionaire, you
[00:24:50] know, and I’m like, “That’s possible,
[00:24:52] but it’s a unicorn.” It’s got to have
[00:24:53] somebody who knows all of these things
[00:24:55] in one person. And it’s really really
[00:24:58] hard to find cuz I’ve hired a lot of
[00:24:59] people in my career and people really
[00:25:01] specialize and it amazes me how
[00:25:03] specialized they are and how I always
[00:25:06] felt that pigeonholes them, you know, I
[00:25:08] in my opinion. So, you usually know
[00:25:10] someone, yeah, I know mobile, I know
[00:25:11] backwards or whatever. I said, well, can
[00:25:13] I talk to you about the database
[00:25:14] structures? Oh, I don’t know about that.
[00:25:16] I just know what I give to a develop
[00:25:18] back-end developer and he tells me the
[00:25:19] fields I got to get, right? So, I think
[00:25:21] we still need those kind of people and I
[00:25:23] guess that’s the thing you know, I had
[00:25:25] to unlearn and realize that we’re not
[00:25:27] going to find all these full-stack
[00:25:28] developers. And that’s before even AI is
[00:25:31] that I had to kind of accept that people
[00:25:34] specialized and still have a fairly
[00:25:36] large team, right? I still need a UX
[00:25:38] designer, I still need a BA, I still
[00:25:40] need a DBA, I still need an architect, I
[00:25:43] still need a security person, I still
[00:25:45] need a DevOps person, right? And you
[00:25:47] know, in theory, yeah, all these could
[00:25:48] be done by one person if they have the
[00:25:50] AI and they have the knowledge, but you
[00:25:52] really rarely find that in any
[00:25:53] individual.
[00:25:54] >> Yeah.
[00:25:55] >> Yeah.
[00:25:55] >> Yeah, it’s kind of a roundabout way to
[00:25:57] get to it, but
[00:25:57] >> So, James, you know, having worked with
[00:25:59] Deloitte, McDonald’s, huge names on your
[00:26:02] portfolio, when and why did you decide
[00:26:04] to leave it all behind and do something
[00:26:07] of your own? What problem were you
[00:26:08] trying to solve?
[00:26:09] >> Well, I mean, I always build something.
[00:26:12] It’s It’s how I stay current. It’s how
[00:26:14] I’m able to face developers in a room
[00:26:15] and they take me quite serious cuz I I
[00:26:18] know all the technologies. It’s always
[00:26:19] funny when a developer tries to kind of
[00:26:21] bluff me into something and I tell him,
[00:26:23] yeah, bring up the code and they’re
[00:26:24] like, what? So, I go, “Dude, I can code
[00:26:27] like with the best of them.” And you
[00:26:29] know, you’re going to show me your code
[00:26:31] and you’re going to prove what you just
[00:26:32] said to me, right? So, I I think that
[00:26:34] having that coding background is is
[00:26:37] really good and I always develop
[00:26:39] something. So, sometimes I work for a
[00:26:41] large company and sometimes I like to
[00:26:43] work for myself and it all depends. And
[00:26:45] once I finish this, I may go back and
[00:26:47] work for another large company again
[00:26:49] because it’s great exposure, you know,
[00:26:51] kind of do this on the side, maybe, or
[00:26:53] do it full-time. It all depends. I mean,
[00:26:54] if it takes off, then yes, I’m going to
[00:26:56] do this full-time. If it doesn’t, then
[00:26:58] well, I go back and work for a large
[00:26:59] company, and I learn a lot from everyone
[00:27:01] around me. So, there’s benefits to both.
[00:27:03] Running your own company, you got the
[00:27:04] freedom. I can explore any technology I
[00:27:06] want. I don’t have to be held by
[00:27:08] people’s deadlines or perceptions of
[00:27:09] what an application is. So, I have a lot
[00:27:11] of freedom to grow my skills. And then
[00:27:13] working for a large corporation, I get
[00:27:15] to benefit from the knowledge of other
[00:27:16] people, look at larger scale things,
[00:27:18] have to work within the constructs of
[00:27:20] what the company is designing, right?
[00:27:22] Sometimes, you know, a little uh forest
[00:27:24] restrictions is you learn a lot from it.
[00:27:26] So, you don’t have the freedom. So, you
[00:27:27] have to follow the rules, and that’s not
[00:27:29] a bad thing. You learn something from
[00:27:31] it.
[00:27:31] >> You know, another aspect that I’d like
[00:27:32] to discuss that, what change have you
[00:27:34] been observing or seeing in in the new
[00:27:37] in the new hires, you know, the um the
[00:27:39] interns or the software engineers in
[00:27:41] comparison, you know, to the ones who’ve
[00:27:43] been recruited way before AI even
[00:27:45] existed. Are they more smart, or they
[00:27:48] just know how to work with AI and lacks
[00:27:51] their own contextual or knowledge depth?
[00:27:54] >> You know, I It depends It depends on
[00:27:57] where they come from. If If you uh if if
[00:27:59] you’re This is really going to upset
[00:28:01] developers. I know this. I’m sorry, but
[00:28:03] it’s my experience. I’ve hired literally
[00:28:05] hundreds and hundreds of developers. If
[00:28:07] If you’re a Python programmer, you’re
[00:28:09] attracted to it because probably someone
[00:28:11] said the phrase, “Oh, man, it’s easy,
[00:28:12] dude. You don’t have to learn all this
[00:28:14] stuff from Java, right?” And I usually
[00:28:16] find they’re not really good developers
[00:28:18] because it’s easy because that’s why
[00:28:21] they went to it, cuz they didn’t want to
[00:28:22] really work with the challenging stuff.
[00:28:24] The best developers I’ve ever seen in my
[00:28:26] life, and they may know Python, but it’s
[00:28:29] the Java developers. These guys are the
[00:28:30] best uh the best of the best. And every
[00:28:33] time I work with them, they amaze me.
[00:28:35] And they know how to use AI, they know
[00:28:37] how to make it sing, and it’s a world of
[00:28:40] difference. It’s like you could hand a
[00:28:42] Stradivarius to somebody, you know, in
[00:28:44] high school that you know, didn’t really
[00:28:45] know you know, lessons very well, and
[00:28:47] the and the instrument sounds okay and
[00:28:49] then you hand it to somebody who is a
[00:28:51] concert violinist and they play that
[00:28:54] thing and it brings tears to your eyes
[00:28:55] and that is the difference of the
[00:28:57] individual using AI. It’s not AI, it’s
[00:29:00] still the person. I’m making AI do
[00:29:03] incredible things because I have
[00:29:05] knowledge in all these different
[00:29:06] backgrounds, mobile, front end, back
[00:29:08] end, database, the whole bit. So, I can
[00:29:10] ask it in ways that other people will
[00:29:12] not ask it in their very narrow silos,
[00:29:15] right? So, I think the Java developers
[00:29:17] are usually more established, older
[00:29:18] developers because that was the language
[00:29:20] of its time and it was hot. So, I would
[00:29:21] say they are way more skilled and more
[00:29:25] more suited for this AI world than like
[00:29:27] the Python developers that are coming
[00:29:29] because they have never had to work with
[00:29:31] anything really challenging and
[00:29:32] difficult. So, and it’s just my opinion,
[00:29:34] right? And I I’m sure I’m wrong. I’m
[00:29:36] sure people would send me a lot of hate
[00:29:38] mail on that, but
[00:29:40] from my experience, from the limited
[00:29:42] developers I’ve worked with, that the
[00:29:44] Java developers have been way better at
[00:29:46] this game than the Python developers.
[00:29:49] And and I could extend this argument
[00:29:51] into the Microsoft world because I come
[00:29:52] from that world, too, right? The Visual
[00:29:54] Basic and the C# developers, those
[00:29:56] things compiled at one point to the same
[00:29:58] exact binary, to- totally different
[00:30:01] programmers. So, I’m sure the Microsoft
[00:30:04] world is facing the same thing. That,
[00:30:06] you know, you put AI in the hands of a
[00:30:08] C# developer compared to an old Visual
[00:30:10] Basic developer, you’re going to get two
[00:30:12] different applications at two different
[00:30:14] skill levels.
[00:30:15] >> Well, James is moving towards the last
[00:30:16] question. What advice or one thing you
[00:30:19] would want to tell to the CTOs or IT
[00:30:21] leaders listening to, you know, get
[00:30:24] right about AI and, you know, if they
[00:30:27] should be trusting AI at this point or
[00:30:28] not?
[00:30:29] >> I would give them this advice and this
[00:30:31] this is really interesting because one
[00:30:33] of the studies is uh that I that I
[00:30:35] posted on my LinkedIn was a study from
[00:30:38] Stanford University and the first thing
[00:30:39] the guy said was CTOs all over the world
[00:30:42] are going crazy cuz their CEOs are
[00:30:44] busting in their office saying, ‘Hey, I
[00:30:47] saw this guy on LinkedIn. He generated
[00:30:49] this application in a half hour. Why the
[00:30:51] hell aren’t we doing it?'” So, you know,
[00:30:52] here’s what I’m going to tell the CEO.
[00:30:54] Okay, so thank you for that, okay? But
[00:30:57] here’s the thing. Nothing on LinkedIn is
[00:30:59] vetted or peer-reviewed, and all of this
[00:31:02] stuff is like hype just to get, you
[00:31:04] know, people to look at them. I don’t
[00:31:06] agree with that statement that you could
[00:31:08] generate it in a half hour, okay? One,
[00:31:10] I’ll tell you I’ll prove it to you why I
[00:31:12] don’t believe that. Because if that were
[00:31:14] true, I wouldn’t be working for you. I
[00:31:16] would be doing it myself, and I would go
[00:31:18] build that in an hour and go sell it,
[00:31:21] and then build something else in an hour
[00:31:22] and go sell it. And I could create a
[00:31:24] fleet of applications and become a
[00:31:25] billionaire overnight. So, if I’m not
[00:31:27] doing that with all my knowledge, then I
[00:31:29] don’t think it can be done. So, that’s
[00:31:32] the problem is there’s this perception
[00:31:34] that you can build stuff in 10 minutes,
[00:31:37] and then you have this massive product.
[00:31:39] only may be true on my platform, but
[00:31:41] that’s not cuz it’s generating code.
[00:31:43] It’s configurating a stable platform
[00:31:45] that already exists. So, that can be
[00:31:48] generated in 10 minutes, but you’re
[00:31:49] building that out of base that’s been
[00:31:52] built over a year by multiple
[00:31:53] developers. So, it’s a different It’s a
[00:31:55] totally different thing than generating
[00:31:57] code. So, I guess the real advice is
[00:31:59] stop believing the hype. Listen to your
[00:32:01] technical leadership and believe them
[00:32:03] because they’re not making it up. What
[00:32:05] Why would they resist doing that? And
[00:32:07] the CEOs of the world out there can
[00:32:09] really struggle with that feedback cuz
[00:32:11] they don’t believe it. They’re saying,
[00:32:12] “Well, I got 12 guys I’ve seen on
[00:32:14] LinkedIn who do this.” They’re not. Look
[00:32:15] into it deep. It’s not peer-reviewed.
[00:32:18] They’re These guys are not billionaires.
[00:32:20] You know, they produced something that
[00:32:21] looks fantastic. I can do that, too. And
[00:32:24] under the hood, it’s a mess.
[00:32:25] >> No, this was a really interesting and
[00:32:27] important point that you raised towards
[00:32:28] the end of the conversation. But yeah,
[00:32:31] you know, that’s the ultimate reality
[00:32:32] and truth around it that yes, he was get
[00:32:34] pretty much hyped around this idea. But
[00:32:36] then they’re also coming, you know, to
[00:32:38] the understanding that most of this is
[00:32:40] just gimmick and it’s not really working
[00:32:43] in reality. And hence, you know, we are
[00:32:45] where we are.
[00:32:47] >> And again, it also really depends on
[00:32:49] whose hands it’s in. That’s why you
[00:32:51] still need your technical staff and they
[00:32:53] need to guide this and those skills do
[00:32:56] not go away cuz AI doesn’t have them.
[00:32:58] Not yet, you know, and I I’m sure I’ll
[00:33:00] get the not yet, but from what I’ve
[00:33:02] seen, it does amazing things on a small
[00:33:04] scale. On large, it doesn’t do so well.
[00:33:07] You know, and and to be fair, if
[00:33:08] somebody came into my office and said,
[00:33:10] “Build me Salesforce.” I’d be like, “Oh
[00:33:12] my god, okay.” I built it and then I
[00:33:14] show it to them and they go, “Well,
[00:33:15] that’s not what I wanted.” Well, your
[00:33:16] requirements were so short that I feel I
[00:33:19] met your requirements, but your
[00:33:21] requirements are not well defined,
[00:33:22] right? And this goes back to our earlier
[00:33:24] discussion about how important now
[00:33:26] requirements are. And the more detail
[00:33:28] you have, the better your AI output will
[00:33:30] be. So, that’s my message.
[00:33:31] >> Yeah, [clears throat] yeah, I know,
[00:33:32] 100%. Well, James, thank you so much,
[00:33:35] you know, for this conversation and for
[00:33:36] your honesty. Absolute pleasure to have
[00:33:38] you on Tech on Hinch with us.
[00:33:39] >> Thank you, Rubina. This has been a great
[00:33:41] pleasure. Thank you for having me.
[00:33:44] >> [music]
