Custom Software Development, AI Integration, & Client Engagement Strategies with Andrew Duncan

[00:00:00] Richtig spannend, richtig vielfältig. Das ist deine Karriere bei Kaufland. Ob Trainee-Programm, Direkteinstieg, Studentenjob oder Praktikum, finde bei uns den Einstieg, der zu dir passt. Profitiere von einer attraktiven Vergütung, spannenden Aufgaben und individuellen Entwicklungsmöglichkeiten. Werde jetzt Teil unseres Teams. Bewirb dich jetzt unter www.kaufland.de.

[00:00:23] slash Studenten. Kaufland. Hier bin ich richtig.

[00:00:55] Thanks for having me.

[00:00:56] So I'm kind of super excited to talk to you because you specialize in software development as a consultancy. You generally work with clients where you're helping them build custom software.

[00:01:07] And what's interesting to me is this is a very complimentary space to those of us that deliver IT services, because oftentimes we're working with customers and we're helping them with their needs, and they will identify some pain point.

[00:01:19] And we're deciding what that process is to get it fixed. We may look for off the shelf software, but we might hit the limits of that and not all the people that deliver the IT services have that expertise. Tell me a little bit about from your perspective when you're working with a customer, like what are the indicators that a company should start working with a software consultant?

[00:01:43] Yeah, that's a really good question. I'm also a big fan of off the shelf software because building software is a time intensive, complicated, expensive process.

[00:01:53] So if you can buy the solution already for $2,200 bucks a month, then you should do that because it's going to be much more expensive to build it.

[00:02:00] But there are times when it's appropriate to build your own software. And one of the first things that I tell clients is, if it doesn't have to do with your secret sauce, the process that you follow to provide value to your clients, then you shouldn't be building it.

[00:02:19] It shouldn't be something that you focus and spend resources on just because it's not going to help you, you know, make better tasting beer, for example. Right.

[00:02:27] So, you know, if it's not a part of your your secret sauce, don't build it by it. There's probably something exists already. Right.

[00:02:34] For example, if you're a payroll invoicing company, then you should probably be building a payroll invoicing software.

[00:02:41] But you don't have to build like a CRM, for example. Right. So that's the first indicator of when you should or should not.

[00:02:48] But let's say, for example, you do have or you believe you have a need for building software.

[00:02:54] I give you a real world example where a client of mine needed to build some software.

[00:02:59] So they they are a, you know, a marketing research firm and part of their offering is curating panels of people that can participate in the research studies.

[00:03:11] And so they need a way to reach the panel when they have a study to run that matches their demographic data.

[00:03:19] And so they decided to build a mobile app that their panelists exclusively for their panelists could download so they could, you know, they could browse the existing research that's available for them to participate in.

[00:03:31] And so they could send them notifications that a new study has come out for them to reach.

[00:03:35] So that's an example of where.

[00:03:39] A custom software solution helped this company deliver more deliver specifically on their secret sauce, which is curating panels that match the research study.

[00:03:48] So that's a big piece.

[00:03:49] Sure. So that makes it that makes some sense.

[00:03:51] Now, what I also want to think about, though, is, is in a world of kind of integrations. Right.

[00:03:56] We always talk about linking software.

[00:03:58] Now, in my mind, there is some element of development that goes along with the linking.

[00:04:02] I don't want to be completely dismissive and say, like, well, that's just four clicks and we're done because I recognize that there is some some art and software development that goes into linking.

[00:04:12] But at the same time, it is simpler to make two systems talk to one another than try and build something, you know, a new custom system to talk to a back end.

[00:04:21] But I'm going to use your example because it makes a lot of sense to me, like my first inclination might have been say, well, well, there's some CRM elements to that.

[00:04:28] Right. And so I would think that there is an element where maybe we want to extend a piece of software and have like a custom front end for the for the customer to give that special sauce element.

[00:04:41] But we want to keep a lot of it off the shelf. Talk to me how you frame that thinking when you're talking to a customer.

[00:04:48] Yeah, great question. And it fits right, right within the wheelhouse of this example, because my client does use a CRM and they do keep their panelists information in that CRM.

[00:05:00] And so, you know, to populate, for example, their mobile app of who is allowed to log in because it's limited to the panelists.

[00:05:11] We linked the database that drives the mobile application and houses the research information to their CRM so that whenever they add a new panelists because they have a separate process that they go through to vet panelists.

[00:05:24] They collect demographic information to figure out which studies they may qualify for and be, you know, that participates in.

[00:05:30] And so we we connected the two systems, right, their custom mobile application that we built and their CRM so that whenever they have moved a panelist through the process of the vetting process, they can automatically feed them right into the app.

[00:05:44] They get an invite to download the app. They can log in and then start participating in the research that's available to them.

[00:05:51] So it is much simpler to link to systems than it is to build a system.

[00:05:56] And like there are tools off the shelf now that you can use to integrate various software systems.

[00:06:01] Zapier is one of the well known ones that you can use to plug in to connect to different APIs, for example.

[00:06:10] Now, most of the time customers don't really know their requirements.

[00:06:15] Now, we can all say this as technical people and just sort of smile because it doesn't matter what you're building or implementing.

[00:06:21] Customers have some sense of their need, but never are particularly good at articulating that.

[00:06:28] When we think about it from like an infrastructure and a security management perspective, they don't know what they don't know.

[00:06:32] I know that's true in software development as well, but I suspect that some of the things they don't know are a little different.

[00:06:39] Talk to me about kind of your experiences on what you have to work through to have a successful engagement with a customer and what you have to discover to get it right.

[00:06:49] Yeah. So one of the big pieces of our process is making sure that whatever solution we implement for our clients is going to move the needle that they expect it to move.

[00:07:00] Right. So we start out with talking about what they hope to achieve. Right.

[00:07:04] It's a concept of the five why's right. Yes.

[00:07:07] But, you know, yes, why you answer? Yes, why you do you repeat that five times.

[00:07:11] You can kind of get to the core of the problem or the desired outcome.

[00:07:15] So we go through that process with our clients to make sure that whatever we're going to recommend is the right approach.

[00:07:21] Sometimes that results in off the shelf software. Sometimes that results in integrations.

[00:07:25] Sometimes that results in custom software. And so we start we after we go through that process of understanding what they hope that desired outcome is.

[00:07:33] We define KPIs that we can use to measure once we launch the solution that we're implementing for our clients.

[00:07:40] So there's a there's an entire industry term discovery process that we go through with our clients to make sure that the money isn't wasted.

[00:07:49] One of the reason I got into this business, one of my biggest peeves is building software that that that people don't use or that doesn't make impact.

[00:07:56] And so I'm a I'm a get stuff done kind of person.

[00:07:59] So I want to make sure that whatever we're doing is is really going to help the business succeed.

[00:08:04] That's a that's a as a technologist. My my my biggest passion is moving the business forward.

[00:08:10] So the obvious next question there is, so how do you measure that?

[00:08:13] How do you measure that you're really making that impact?

[00:08:15] Yeah, that's a great question. So, you know, it depends.

[00:08:18] It's a it's business and project specific. Right.

[00:08:20] In the example of the mobile app that I talked about, it was how many of the total panelists engage with the mobile app, both download it.

[00:08:27] They use it regularly. So monthly active users, which we're all familiar with, our Spotify and Netflix subscriptions.

[00:08:34] And then also how many of them are actually engaging in the in the research, the research that's available for them to engage in.

[00:08:40] So those are metrics that we we can define. We know we know the total number of of panelists that are available.

[00:08:46] So we know how many are engaging with the mobile app and the content therein.

[00:08:51] And we set some expected engagement rates that we hope to hit when we launch.

[00:08:57] And then over time, as we communicate the launch of this, we can determine who is who is engaging.

[00:09:03] And then you can implement things like retargeting campaigns for these panelists.

[00:09:06] We didn't do this for this particular project, but because engagement was actually quite high from the launch, which was awesome.

[00:09:11] But you can implement retargeting campaigns to help those metrics come along, again, assuming that you are already implementing the right solution.

[00:09:19] Now, start at most startups and SMBs are, you know, historically very resource constrained.

[00:09:26] Right. And it's a nice way of saying they don't have as much money sometimes.

[00:09:30] And they are very, very careful about that.

[00:09:33] Now, software development, like you're talking about, is very bespoke.

[00:09:36] Right. And and generally is done in a bespoke style of engagement.

[00:09:41] Lots of business models and I'm not making an assumption here, but it's generally done in a way where it's project based, which is it tends to be a little bit more on the transactional side than like a fixed fee.

[00:09:51] We can do anything without worrying too much about the contract venue itself.

[00:09:56] Talk to me about how you are making sure to leverage that smaller level of resource and spend effectively.

[00:10:04] Yeah. So this is something I'm I guess I'm quite proud of because software engineering is is bespoke.

[00:10:12] Right. You're building something custom from the ground up.

[00:10:15] There are a number of ways that a project can go off track.

[00:10:19] So there are a couple of things that we do at Vertice Labs that help mitigate that.

[00:10:23] So startups and SMBs are our target market.

[00:10:26] I love the startup space. I love to organize chaos.

[00:10:29] I love laying out the I call it jumping from lily pad to lily pad across the lake to make sure because at any point you could fall in and that the company doesn't exist anymore for startup.

[00:10:39] Right. So there's a couple of things that we do.

[00:10:42] First, our process is pretty standardized.

[00:10:44] So we go through the discovery phase that I talked about earlier to make sure we're building the right thing and we're able to measure it.

[00:10:52] We have a standardized set of technologies that we offer.

[00:10:56] And when I say technologies, I mean not languages or clouds necessarily, but the way that we approach implementing those languages and clouds.

[00:11:05] Right. And so we're able to reuse a lot of that from project to projects.

[00:11:08] We have starter templates and that that range from anything from a very simple mobile app to very complex enterprise level type software.

[00:11:17] And so we're able to bring those to every new client that we encounter.

[00:11:20] And then one of the big things that we do is really early on in the software development process, we we create an environment for our clients to use because software is so bespoke and solving a specific problem for the client.

[00:11:35] They have to be pretty heavily involved in the process of developing, developing it because they know their business and they know their expected outcome as an outside consultant coming in.

[00:11:45] We drink from the fire hose, learn as fast as we can about the business to make sure that we're building it correctly.

[00:11:51] But that happens by by deep engagement with the client.

[00:11:55] So we have, you know, we have, you know, my standard process is three meetings a week where we're really talking about, especially very early on in a project lifecycle where we're really talking about what it needs to be.

[00:12:06] And that's not, you know, there's there's very simple.

[00:12:08] There's not very simple, but standard things like authentication and authorization, roles and permissions and things like that.

[00:12:14] Storing files, you know, taking payment that is common in software that we've all done 100 times.

[00:12:20] We don't need to talk about that, but we really need to talk about is what is it that the business process looks like that we are trying to improve, right?

[00:12:28] If we're doing business processes or what is it about the customer and then that's the experience that we need to get right to make sure that the product is going to hit the mark.

[00:12:36] So, you know, getting that quick feedback in the development process all the way down to the, you know, the entry level developer who may be working on the project is important because as developers are making hundreds of micro decisions a day when they're writing code, the more ingrained in the business context that they are, the better it is that they can make assumptions on those decisions.

[00:13:01] So I think that helps a lot.

[00:13:02] I see that go wrong a lot of the times in different projects for industry companies where, you know, they're like, go build this thing, but we don't really want to talk that much to you, right?

[00:13:14] Like that's problematic because something will get built, but it's probably not what you want to have built.

[00:13:19] Now, listeners are going to hear a lot of thematic stuff that sounds very similar to the way we approach things in IT services and managed services.

[00:13:26] So I'm glad I asked that question that way.

[00:13:28] Now, I want to ask, you know, I'm not going to say anything in particular on these because I want to get your insight.

[00:13:34] Like what are the sort of trends in custom software development that you're watching for that are influential right now?

[00:13:43] Oh, yes.

[00:13:44] Such a great question.

[00:13:45] Well, let me start with AI.

[00:13:47] I figured that would be the answer, but I didn't want to lead you there.

[00:13:51] Sure.

[00:13:51] No, it's huge right now, right?

[00:13:53] It's just kind of taken the world by storm when it first came out in 2022, and I am increasingly finding more applications for it with every new project that I do.

[00:14:05] And I've got, you know, I've got five or six going on right now that are including AI and some that are not, but it's transformational for businesses.

[00:14:14] So I'll give you a couple of real world applications of how we've used it at Virtus Labs.

[00:14:20] Personally, I've used it to help with learning my brand voice and learning my business and helping me generate content because, you know, I want to spread the word of what we're doing and offering to the market.

[00:14:30] And I've taught clients how to do that as well.

[00:14:32] But in the software development world, it can do a number of things that were traditionally human and time intensive and typically error prone processes.

[00:14:43] So think like, you know, an insurance company who's processing thousands of claims or invoices every month, right?

[00:14:50] And they're just disparate, right?

[00:14:51] They're different.

[00:14:52] They're different formats from different vendors or contractors.

[00:14:56] And so that's a process now that can be automated.

[00:14:59] AI acts like a human.

[00:15:02] It can read a document.

[00:15:03] You can ask it questions and it can say, well, I think this is what the document says.

[00:15:07] I think these are the data elements in these documents.

[00:15:10] And so you can use it to take that out and feed it into a traditional software program, to your database, to be processed later by department.

[00:15:20] You can use it to reason about things.

[00:15:22] So another good real world use case for a current client of mine is we're using it to automate the asset management of your home, right?

[00:15:31] And so you go around your home, you take pictures of your monitors or your TV or your hot water heater of the manufacturer labels.

[00:15:39] And it can, you know, it'll pull out, make model manufacturer that the year it was manufactured, expect the end of life, the manuals that are associated with it.

[00:15:47] And on the other side of that, for the vendors who repair these items, right, the traditional approach is, you know, I cover these brands and I'll bring a manual up and I'll search the manual.

[00:15:58] Well, you don't have to do that anymore because you can feed all of the manuals into AI.

[00:16:02] And then you can literally say, here's what's happening to, for example, the hot water heater.

[00:16:08] What are the steps to diagnose and fix the problem?

[00:16:12] And because AI has the context of that manual and knows about it intimately, it can give you very clear steps on what you should do next in your diagnostic and repair process.

[00:16:24] So it's quite powerful.

[00:16:26] Those are like two, I would say, simple but very powerful use cases for it.

[00:16:30] Now, I love those use cases and they're thematically very similar to ones I've been talking about on the show.

[00:16:35] But notable was the one you didn't lead with, which I would have expected from a custom software developer.

[00:16:41] And that's the generation of code.

[00:16:43] Talk to me a little bit about that.

[00:16:45] And you're smiling, as we said, for those that are catching just the audio version of this, like you're smiling.

[00:16:50] Tell me a little bit about that.

[00:16:51] Is that you're not doing it there or are you using it in a different way?

[00:16:54] Tell me how about it impacts code generation.

[00:16:57] Yeah, we definitely are using it there.

[00:16:59] If you're not, you're behind the curve already.

[00:17:02] So I use it.

[00:17:03] So there's a tool called Cursor, which is an IDE, like Visual Studio Code or Visual Studio or Eclipse, depending on what language you're writing.

[00:17:09] And it has basically the common LLMs built in.

[00:17:15] And so it has the context of your entire code base and where you're writing code.

[00:17:19] So my favorite thing to do if I'm writing code in there is to tell it what to do as if I was explaining it to another developer.

[00:17:27] And it will go and do it.

[00:17:28] And it gets it wrong sometimes.

[00:17:30] Sometimes it gets it perfectly right.

[00:17:31] Right. And you just you iterate.

[00:17:34] You just tweak it.

[00:17:34] So you're like it produces some code.

[00:17:36] It is not quite right.

[00:17:37] Or you need to or you need to tweak it just a little bit to cover another use case maybe that you didn't provide it to start.

[00:17:42] But I would say it gets somewhere between 80 and 90 percent there where the cognitive load on the software engineer is significantly reduced, which gives them the ability to hold more of the business context in their brain for the problem they're trying to solve.

[00:17:57] So I think it significantly increases the efficiency of the software development process.

[00:18:00] The other aspect of that is this is my favorite use case that I haven't really talked much about is is using the LLMs to write the code in real time at runtime.

[00:18:13] So that is you write some traditional software and in that traditional software, you deploy with it the the AI capabilities.

[00:18:23] And when your software runs, it will write code on the fly to solve a problem that you need to solve.

[00:18:29] And then we'll execute that code at the same time.

[00:18:31] So here's here's like a practical use case of that, that a buddy of mine is using right now in his business.

[00:18:36] So he is doing a lot of API integration, bespoke API integrations for a bunch of different vendors that they're integrating with.

[00:18:46] And instead of traditionally, you would you would sit down, read an API spec doc, write the code and deploy it, deploy it, test it and then iterate in that manner.

[00:18:55] Right. And so that's days.

[00:18:57] It's time consuming.

[00:18:57] I mean, the the process that he follows now is he will tell the AI model, hey, you're building an AI API integration.

[00:19:08] Here's the general spec because he has the specs.

[00:19:11] Write the code to to to integrate with this API, run it, test it, get the error back.

[00:19:17] If an error occurs or if a success occurs is done, the error comes back from the API.

[00:19:21] If he feeds the error into the model and says, here's here's the error.

[00:19:26] Will you fix it?

[00:19:27] It updates the code and then redeploys it.

[00:19:30] And after it runs a few times, eventually you have a functioning integration.

[00:19:35] And then he scaled that across hundreds of integrations.

[00:19:38] And it's all running in real time at runtime.

[00:19:41] And that's so that's really, really cool application.

[00:19:44] I haven't done myself, but I've seen.

[00:19:46] So I'm intrigued by that because one of the things that we think about in software is, is, of course, we want it to be deterministic.

[00:19:52] Right. That we wanted to actually to know every single time it will behave the same way and we will get the same reliable result.

[00:19:58] And one of the things about a lot of these particularly generative AI based models is, of course, they're probabilistic.

[00:20:05] Right. So there.

[00:20:06] So you're gambling, essentially, that it will end up in the right place.

[00:20:11] I'm using a slightly negative term to just to make it distinct.

[00:20:15] Do you have particular concerns about using from straying away from a methodology that is so defined?

[00:20:24] And what are those, you know, as you think about the fact that, well, it may not always work the same way every single time.

[00:20:32] So, no, I don't.

[00:20:36] I don't have reservations about this.

[00:20:37] And here's why.

[00:20:38] It's because, you know, I've led large teams.

[00:20:40] I've led, you know, a team of one, which is myself.

[00:20:42] And I'm a human.

[00:20:43] My teams are humans.

[00:20:46] And humans are non-deterministic.

[00:20:48] Right.

[00:20:48] Or I like your term.

[00:20:50] It's a probabilistic.

[00:20:51] I hadn't heard before.

[00:20:52] That was interesting.

[00:20:53] And so I'm not worried about that because we are already living in the world that operates that way.

[00:20:58] Right.

[00:20:58] We're surrounded by a bunch of humans.

[00:20:59] Now, we have some reasonable expectation on how, you know, my friends will behave.

[00:21:05] My wife will behave.

[00:21:06] My kids will behave because we have a relationship with them.

[00:21:08] Right.

[00:21:08] Right.

[00:21:09] You don't necessarily have that with AI, but you do still have a reasonable expectation on how it will behave based on, you know, based on some of the iterative approaches you've used when you're building it out to solve the problem that you're solving.

[00:21:22] So it doesn't give me worry.

[00:21:25] It gives me, I would say, I think about it in a way that when you apply it, you solve the 80% use case.

[00:21:39] And I always call it human in the loop where you need to allow for your software process that is using AI to allow human interaction when it gets stuck or it gets something wrong because that will happen from time to time.

[00:21:52] So that's how you get, that's how you overcome that.

[00:21:54] You deploy a, you know, an AI agent or, you know, some AI software process that's embedded in your software process now and you give it an exit hatch.

[00:22:05] Gotcha.

[00:22:06] Well, Andrew Duncan is the founder of Vertice Labs, a boutique software engineering firm specializing in custom software delivery and strategic technology advisory services.

[00:22:15] Duncan brings a wealth of experience in application development, vendor management, and consulting for both major enterprises and startups to help organizations launch and scale products.

[00:22:25] Duncan, this has been fascinating.

[00:22:26] If people are interested in learning more and reaching out, what's the best way to do that?

[00:22:30] Yeah, absolutely.

[00:22:30] You can go to verticelabs.io or email me at duncan at verticelabs.io.

[00:22:36] Awesome.

[00:22:37] Thanks for joining me.

[00:22:38] Thanks for having me.

[00:22:40] Are you ready to get your brand in front of the tech leaders shaping the future of managed services?

[00:22:46] Here at the Business of Tech, we offer flexible sponsorship opportunities to meet your needs, whether it's live show sponsorship, podcast advertising, event promotion, or custom webinars.

[00:22:57] From affordable exposure options to exclusive sponsorships, our offerings are designed to fit businesses and vendors of all sizes looking to make an impact.

[00:23:07] Prices start at just $500 per month, making our packages a fraction of typical event sponsorship costs.

[00:23:16] Be a part of the conversation that matters to IT service providers worldwide.

[00:23:22] Join us at MSP Radio and amplify your message where it counts.

[00:23:26] Visit MSP Radio dot com slash engage today to explore all the ways we can help you grow.

[00:23:34] The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines posted at businessof.tech.

[00:23:44] If you like the content, please make sure to hit that like button and follow or subscribe.

[00:23:49] It's free and easy and the best way to support the show and help us grow.

[00:23:53] You can also check out our Patreon where you can join the Business of Tech community at patreon.com slash MSP Radio or buy our Why Do We Care merch at businessof.tech.

[00:24:07] Finally, if you're interested in advertising on this show, visit MSP Radio dot com slash engage.

[00:24:13] Once again, thanks for listening to me and I will talk to you again on our next episode of the Business of Tech.

[00:24:22] Part of the MSP Radio Network.

[00:24:25] Richtig spannend, richtig vielfältig.

[00:24:28] Das ist deine Karriere bei Kaufland.

[00:24:30] Ob Trainee-Programm, Direkteinstieg, Studentenjob oder Praktikum.

[00:24:35] Finde bei uns den Einstieg, der zu dir passt.

[00:24:37] Profitiere von einer attraktiven Vergütung, spannenden Aufgaben und individuellen Entwicklungsmöglichkeiten.

[00:24:44] Werde jetzt Teil unseres Teams.

[00:24:46] Bewirb dich jetzt unter kaufland.de slash studenten.

[00:24:50] Kaufland.

[00:24:51] Hier bin ich richtig.