Build or Buy: Is It Cheaper to Vibe Code with AI or Use SaaS?

Welcome to the Ecommerce Finance Podcast. I'm Stephen Brown with LedgerGurus. Today I have Stacy Walker, Director of Growth here at LedgerGurus, and we're going to discuss build or buy. Is it cheaper to vibe code with AI or use SaaS?

Stacy, welcome back.

0:35 Stacy Walker

Thanks for having me, Stephen.

0:37 Stephen Brown

In our previous episode we talked a little bit about AI costs. And along that line I want to talk about vibe coding, because it seems like everybody's talking about it, and there's all this conversation around the death of software. We're all just going to build our own tools, and all the SaaS companies can go away, and we'll just send all that money to OpenAI and Anthropic. Problem solved.

0:58 Stacy Walker

Yeah, I want to see how that works out.

1:02 Stephen Brown

Me too.

So I want to explore this, because I think ecommerce businesses are very technically savvy. They're using a lot of technology. And I'm hearing about a lot of operators who are vibe coding apps and believing that it's the answer to all their problems.

What do you think, Stacy? Are you going to go vibe code all your stack? Get rid of your CRM and just build your own?

1:24 Stacy Walker

No. I just don't see a world where it's going to work. I just don't. There are too many things that change too often. I don't see how it's going to happen. Talk me out of it, though. Maybe I'm wrong.

1:41 Stephen Brown

Well, let's talk to the implications of building software, and maybe end up on where it might make sense and where it doesn't.

As a little background, I spent 20 years in enterprise software, so I feel like I can talk about this from experience. I worked for software companies from 1997 to 2016, including a Fortune 500 and a couple of publicly traded companies.

2:10 Stephen Brown

So I've been around building software for a long time. And obviously this changes in the age of AI.

I was not a great software developer. I got into other aspects of the software industry. But I look at AI tools and I'm like, hey, I might have been a half-decent software developer, because I'm pretty good with the logic and how things should work. But I always struggled with the syntax, and which function do I use for this, and how to use it. It was not my forte.

And so I can see why people are like, I can do it. And I've built some stuff now with AI tools, and I'm like, man, I can build things, and it's awesome. But it's easier said than done.

3:06 What people are trying to build

Stacy Walker

Yeah, for real. So tell me what you see the most with people vibe coding. I'm curious what you're seeing in your role, and then I'll tell you what I'm seeing in mine.

3:06 Stephen Brown

I'm hearing a lot of stuff. I'm hearing people say, I'm going to build my own inventory management system. Or they're saying, I'm going to go replace that Shopify plugin. And that's mostly the ecommerce space.

I haven't heard anybody say they're going to replace Shopify or a Klaviyo, one of these big systems. But I would consider an inventory management system a pretty serious system.

3:35 Stephen Brown

And I know in the accounting space there are conversations around, could I vibe code my own accounting software, because all these accounting software vendors are terrible and I'm just going to build my own?

And that's what caused me to want to have this conversation. Because I'm like, be careful. You might not know what you're getting into.

3:52 Stacy Walker

Yeah. So what do you think the implications are?

I'm thinking about the inventory management systems that people are trying to vibe code. But you brought up something I'm curious about. You said they're dropping plugins on Shopify and building their own. So are there security issues with that?

4:12 Stephen Brown

There are a number of issues, and maybe we should walk through those. Security is one of them.

4:16 Stacy Walker

Okay, let's do it.

4:18 The cost of building and maintaining

Stephen Brown

So I would do a callback to our previous episode first.

Depending on what you're building and how much you're using AI in your software, one of the things you've got to think about is tokens and subscriptions and how much consumption that is. What is the cost of building that software?

As you and I talked about previously, there are hard costs going into AI software that you and I believe are going to become a bigger issue. So first, how much is it going to cost to build that software? Right now it's not too expensive. But I think as you're getting into Claude Code and the other tools, those costs are starting to become usage-based. And so you've got to think about how much it's going to actually cost to build that.

5:06 Stacy Walker

Yeah. And how much is it going to cost to keep it running, right? What happens when your process changes?

5:12 Stephen Brown

That's the second point, right?

A lot of times people think about software and they're like, I built this cool app. But software is not a one-and-done thing. It's a living, breathing thing that you're constantly changing. You'll be like, I need to fix this, or I need this new feature. It's constantly evolving.

So it's more than that upfront cost. There's the ongoing maintenance cost of software, which is why there are subscription fees. Even before that, you had maintenance fees on software, because the vendors were constantly fixing and updating it. You need to think about the ongoing costs of updating that software.

5:55 Stacy Walker

So I think about some of the software we know our clients are using. Think about a sales tax software, where people say, I just set it and forget it, I don't think about it again. But we know that things break.

There are very few things, I can't think of anything in our ecosystem, that you can set and forget as far as software's concerned. So who is so confident with their vibe coding that they feel like they can set it and forget it?

6:28 Stephen Brown

Yeah, I don't know. I would hope people aren't.

And that's the first consideration. APIs change, if you're connecting to Shopify or Amazon or whatever platform. Most software doesn't live in isolation anymore, it's usually connecting to other systems. So you've got to consider that APIs are going to change. There are going to be features you're going to want to add to your tool. And so I think you have to think a little bit about how much you're willing to invest to maintain this software.

6:59 The builder conundrum

Stephen Brown

Along that line, I think there's another thing, and that's the builder conundrum.

7:04 Stacy Walker

What is the builder conundrum?

7:06 Stephen Brown

Not everybody's going to vibe code. Have you vibe coded anything, Stacy?

7:09 Stacy Walker

No, definitely not. I have other people that do it. I don't have to.

7:12 Stephen Brown

Shame on you. How do you even wake up in the morning and look at yourself in the mirror?

But I don't know, when you go to conferences or listen to any podcast, you feel like, my gosh, I should be vibe coding. Everybody seems to be talking about this.

7:27 Stacy Walker

No, because I don't feel like it can be successful long term. And so I'm not putting my efforts there.

7:34 Stephen Brown

Maybe as a builder, somebody who's built software, I feel this guilt, like I'm not vibe coding enough. I have the skills. Okay, I have vibe coding guilt. I'm like, I should be building more software. This is what I did for twenty years. Why am I not building more software?

7:55 Stephen Brown

Not everybody's going to build. There's a skill set to building software, whether you're using old-school integrated development environments or no-code applications. There's an understanding of underlying logic and architectures that you have to develop for yourself.

And so I think this idea that everybody's going to be a software builder is a little bit extreme. Now, I think it'll be easier to create no-code applications. But full-blown applications that are integrating with an API, have a back-end database, a system of record, that's not going to be everybody.

8:35 Stephen Brown

So let's say you go out and decide to build one of those things, or you have somebody on your team who is the builder. That person is now the maintainer of the software. What happens if that person decides to leave? Who's going to maintain that critical software you built?

Or if you're the owner, like, I'm just going to do it myself, are you now a slave to that software? You've got to constantly update it.

9:00 Stephen Brown

And so the builder conundrum is something where, custom software's been around forever. It's not like companies couldn't build their own software, it's just gotten a lot easier. But the builder conundrum doesn't go away. You're still going to need the person, the expert who understands that software, to maintain it, or bring in other people to maintain it. So that's something people have to consider.

9:26 Stacy Walker

Does the vibe coded software always work?

9:28 Stephen Brown

Software breaks. It breaks all the time.

When I was building software we were constantly updating it. You'd find bugs. Like I said, things would change. Software is never static. And so you've got to be constantly updating it. Even if you're not adding features, there's something you've got to do to keep it maintained and up to date.

9:50 Stacy Walker

And does the builder always even know what they need to do? Can they always tell, if something's not working, why it's not working?

9:57 Stephen Brown

Hopefully. In the age of AI I'm hopeful that troubleshooting is easier. I know when I've built some stuff it's way easier, and you can be like, hey, this isn't working.

But I find that because I have a software background, I know how to talk using software terms. I can very much say, hey, do this thing.

10:25 Stephen Brown

I think with no-code software that's a little bit different, because you're building on top of no-code platforms. Obviously you can do things in Claude, and it's probably going to get better. But again, you're going to have to understand how to build things in Claude, and how those artifacts work, or whatever your AI tool of choice is.

So that's problem number one, the builder conundrum. You're going to have to have somebody maintain it, and what happens if that somebody leaves?

10:55 Stephen Brown

Versus a SaaS tool, where that's their problem to solve.

To me it's the analogy of outsourced accounting, which we do, versus in-house. With outsourced, if somebody leaves us, it's our problem to solve. They're paying us to solve a problem. And if the person who was supporting our customer leaves, we have to figure out a solution. Same thing with custom-grown software, you have to figure out how to maintain that. The software vendors will do that for you.

11:27 Stacy Walker

And if you're building your own, you have to be able to know exactly what you want the software to do, and you have to document exactly what the requirements are, right? Whereas if it's a SaaS product, they're doing all of that. So a lot of the heavy lifting, I would say, falls on them. Something breaks, you're going to them.

11:50 Stephen Brown

Yeah. Now, the pro to vibe coding is you can do whatever you want. You don't have to wait for the software vendor to build something, because there are always gaps. You're always like, I really would like it to do this.

Maybe you build a relationship with the vendor, and you're able to get your request onto their backlog and onto their roadmap, and they'll build it for you. The upside of vibe coding is you get to build whatever you want, whenever you want. You can just go ahead and build. So that's a pro to building your own. You get to control what's in the product, and if you need something you can go build it.

12:27 Stacy Walker

And if you go build it, hopefully you have your process so well aligned that you don't miss any steps in the building.

12:34 Stephen Brown

Yeah. I mean, ideally, you're either going to be good at building vibe coded software or you're not, and you'll know that pretty quickly. I think where it's going to get tricky is those who can build something but aren't thinking about some of the other issues that go into software.

12:51 Security and compliance

Stacy Walker

So, like security.

12:51 Stephen Brown

Let's talk about security.

I actually worked on not just software, but information security software. There are a lot of vulnerabilities in software. There's a thesis that vibe coded software, that AI, is going to build perfect software. I don't know if I believe that. I think it'll build better, but I still think there are going to be potential bugs and security vulnerabilities that get created and need to be remediated.

So, how do you know that your software is secure?

13:30 Stephen Brown

The other thing is, when you're interacting with other systems you have to think about authentication, you have to think about how you're using API keys, and you have to think about what type of data is in your system. Is it secure? Is this information that contains customer data, or is it software that's just solving a generic problem where the data's not really sensitive?

Where is that software hosted? How exposed is it? Can other people get to it? I highly doubt people are going to vibe code desktop software, they're probably going to be vibe coding SaaS software. So what are the controls and the security around that software?

These are all things we had to think about in enterprise software. Are you going to think about it? Or are you just going to chase features?

14:23 Stacy Walker

Yeah, exactly. So, Stephen, what about compliance? If I'm vibe coding something, what should I be considering?

14:30 Stephen Brown

There are a lot of regulations around sensitive information. This is hand in hand with security. But if you're handling credit cards or customer records, there are all sorts of things you're supposed to be compliant with, and you may not even realize those things.

A good software company, hopefully your vendors are good, they're going to think about those things. But you could find yourself, especially if your software was not handling that information correctly, with a big lawsuit, or regulators that come down and cause all sorts of nightmares for you.

15:10 Stephen Brown

So that's the other thing around compliance. There are all sorts of laws around the handling of information that you need to be considering.

So that's a big flag. What's in your software? Are you handling any sort of customer information or payment information? Huge flag. That stuff has to be really handled right.

15:37 When building does make sense

Stacy Walker

Must be locked down, for sure. So do you feel like there's ever a time when vibe coding, building your own system, makes sense?

15:46 Stephen Brown

I think so.

Let's take the example of Shopify plugins. Some of these plugins are just basic features that nobody else is solving. Really, they should just be part of a bigger piece of software, but they haven't been absorbed into a bigger piece of software, or the bigger software suites haven't done that. So that could be a good example.

Another example could be, I've got a problem that nobody's solving. We have built some software here at LedgerGurus where we just couldn't find an answer. I spent literally some years trying to find solutions and I just couldn't find it. So finally we just decided to build our own.

16:29 Stephen Brown

I think if the scope is narrow and somewhat stable, it makes sense. If you're building something complex, something like a system of record, I would be cautious, because you may not realize what you're getting into.

The other thing I think about is, are you a software company, or are you a seller of widgets?

16:56 Stephen Brown

And I do think you're going to see some people have success building their own software.

But I'll give you a good example. You and I were at a conference a little while ago, and one of the vendors was on a panel about AI software. He had a lot of developers. He was talking about how they decided to replace a tool with something they built internally, and he was like, it ended up being more expensive.

17:09 Stacy Walker

Oh wow, really?

17:24 Stephen Brown

He's like, yeah. Because I think when you get into tokens, and the cost of building, and the cost of maintaining. He didn't really get into the details, but I think it's all those things that I'm talking about. That stuff adds up.

And so I think you have to be really thoughtful, especially when you're trying to replace software. Is it really cheaper?

17:51 Stephen Brown

Now, if nobody's solving that problem, the answer might be that it makes sense. But if you're like, hey, I'm going to do it better than that XYZ Shopify plugin, maybe you will. Or maybe that was a really good price that you didn't realize.

And heaven forbid, don't go try to build your own Shopify or your own Klaviyo. You're asking for trouble.

18:05 Stacy Walker

Yeah, without a doubt.

I was talking to somebody a couple of days ago at another conference, and he mentioned that he had vibe coded a whole program for his firm, and he did it solely to think through every single piece of the process. Then he handed that over to his developer and said, this is exactly what I need it to do. Because then they could think of everything he hadn't thought of. The security, the compliance, all of those things.

And he felt like it was a great use of his time, because he knew exactly what he wanted once he had gone through that process. What are your thoughts on that?

18:50 Stephen Brown

Again, I think it could make sense, if there's nothing out there.

So part of what I think about is, is the tool internal? Does no solution exist? Do you have that technical continuity, so whoever's building it is going to be around, or you can easily replace them? Is the scope somewhat narrow? And the cost differential, all the costs.

19:21 Stephen Brown

This is where we think, as accountants and CFOs, not just about that subscription cost. You're like, well, I don't have to pay that subscription. Yeah, but maybe you're paying a dude that's way more per month to maintain that software. Or a dudette, right? You don't want to be gender biased, anybody could build that.

On the other side, there are some flags where I would just be like, you really should consider a SaaS tool.

19:30 Stacy Walker

Okay, so let's talk about that.

I am naturally a DIYer. I enjoy the process. But there are things where I'm like, it's not worth the opportunity cost, it's not worth my time, it's not worth my money to try and figure this out. This is a job for a professional. So when do you think that's the case with this?

19:50 Stephen Brown

So coming from our world, I would say number one, is it touching anything money, any sort of credit card records? I'd probably say don't do that.

If it's customer-facing, you'd better be darn sure about its reliability and its usability. If there are any sort of compliance things related to the software, data privacy, payments, going back to that earlier point. And if you don't have confidence in its maintainability, either by you or the person that's building it, then be cautious.

20:40 Stephen Brown

And again, if you're just trying to replace a SaaS software, I'd really look closely and be like, who's likely to be here in five years? Them, or me, or the builder who's building this software? Can I build as quickly as they can build? And as you go down the path of adding more capabilities, who's going to do a better job?

21:15 A build or buy framework, and what happens next

Stacy Walker

That's a really good thing to think about. I like that. So if you had to build a framework for deciding pros and cons, walk us through what that would look like. What would the matrix look like for that?

21:29 Stephen Brown

Okay, this will kind of be a recap of what I just talked about.

Does it touch sensitive data? Buy. Is there SaaS software that's cheaper than what you're going to do? Buy.

What is the technical continuity? Now, that can be an interesting thing, because a lot of startups might not be here in five years. So if you have a lot of confidence in your builder being around, maybe build.

22:00 Stephen Brown

How much is it going to change? How much support is it going to need? Is it a simple tool that's not going to need a lot of support? I could argue build. If there's a ton of complexity, if I'm building on APIs that are constantly changing, you may want to buy.

How core is this to your operation? Now here's the other one, and this is a really good one. If this software goes down, what does that do to your business? And who's picking up the call from the team? That should give some pause about how critical it is.

22:44 Stephen Brown

Now, this is something you can consider as you're working with software vendors. How good is their support? This is something I'd always look at with a software vendor. What's their reputation for support? How do you get support? Because if it's really critical to your business, you want to be able to connect with them.

Now, unfortunately, the software industry has moved away from phone-based support, because it's expensive, toward a lot of chat agents. But still, is there some way to get hold of somebody and escalate? And the more critical the software, the more timely the support that's needed.

23:15 Stephen Brown

So that's what I'd think about. What is the criticality of this software? If it stops working, how does that impact the business? If this stops orders and sales.

And let's say you've got Joe Developer. He's on board, you've locked him up, you've got a really good salary. Joe's on vacation. Maybe he's on a backpacking trip. Nobody can reach him. What happens then?

23:45 Stephen Brown

So I don't know, I might be coming off as a little bit pessimistic toward vibe coding solutions, probably because I've built a crapload of software and I've dealt with all these issues.

That said, we have built tools internally here at LedgerGurus. There's this nervousness that I have, but a lot of those tools I'm building because nobody else is building those things. I've seen a need that nobody else is solving. And so I've been okay going down that path.

24:07 Stephen Brown

So as people are going down the path of build versus buy, I think they need to think about all these considerations, because you can easily make a bad decision.

Now, do you want to hear what I think's going to happen in the next few years?

24:30 Stacy Walker

I always want to hear what you think's going to happen.

24:33 Stephen Brown

I think you're going to see a lot of people that vibe code apps and will probably reverse course and go back to paid software.

24:43 Stacy Walker

I agree.

24:44 Stephen Brown

Now, the bigger thing I think people should be thinking about is not, should I be building my own or buying? I think the question now is, what am I buying?

I think the vendors we're using today are either going to have to level up, or you're going to see a new class of software vendors emerge that are AI-native, that are able to build faster, that are able to use AI features more effectively. And that's probably the next generation of software.

25:15 Stephen Brown

In my career I've seen software evolution. When I started, if I go way back, before I even got into working in software, you and I started on DOS and command-line software. Then it moved to Windows, then we moved to internet-based software, and then we moved to mobile.

Now I not only need to be able to do my software from a browser, but I also want to be able to do it from a mobile app. And the software's the same. And now we're in this age of AI, where it's AI features and capabilities.

26:00 Stephen Brown

And each evolution, you'll see a lot of changes in who the dominant players are. It's often the newer players of software that overtake the older ones. It's the classic innovator's dilemma. Occasionally you'll see legacy vendors that are able to make the leap. That's super hard to do.

So in my mind, I would be advising, don't spend as much time thinking about what you can build by yourself, and think about what vendors are building, the next generation of software, and look at them.

26:44 Stacy Walker

That's great, I love that. So, Stephen, what are your final words of advice?

26:48 Stephen Brown

I would just go through that framework. I'd be cautious about what you're building. I would be very much looking at next-generation software, keeping a close eye on that, building where there are gaps, making sure that it's supportable and sustainable.

Go ahead and hate me, I know all the talking heads are telling you to build your own software. I would just say, as somebody who spent two decades in software, who lived through the web, the mobile, and the cloud revolution, there's a lot to consider. I won't say don't do it. Just do it wisely.

27:27 Stacy Walker

All right. You heard it here, folks.

27:30 Stephen Brown

All right, we'll call that a wrap. Thanks for joining me again, Stacy.

27:33 Stacy Walker

Thanks, Stephen.