Why Agentic AI Will Collapse Thin SaaS and Reshape Enterprise Software

Hey everyone. Welcome back here to Tech Drunk tv. I've got a great guess.

I've been looking forward since I originally scheduled this one to talk to 'em about, 'cause it's an scenario I've been writing and talking about it too. I want to introduce you to Don Murray. Don is the co-founder and CEO of a company called Safe Software.

Hey Don, welcome to Tech Drunk tv. It's great to have you on here. Yeah, thank you so much.

I'm really excited to be here. Good, good, good. So Don, as I mentioned that you're a co-founder, CEO over at Safe Software.

But you know, give us, you weren't born there. What, what have you been doing up, you know, until you've co-founded Safe Software? Yeah.

Yeah. The first thing I, that's probably relevant is I did my master's degree at Simon Frazier University in parallel processing. And, uh, and then, um, after that I went and worked at, uh, McDonald Detweiler and Associates who built ground stations, and I did some work on airspace modeling and where I met the other founder, Dale Lutz.

And, um, and then we, uh, a few years later we founded Safe Software, which is a company that is focused on enabling organizations to work with all kinds of data. Our, sort of, our specialty is anything that's spatial. So IE for companies who have infrastructure on the ground, they obviously care where it is and they want to be able to analyze it and do all sorts of operations and connect it to the rest of their systems.

And so that's a little bit about, about me and what, what Safe Software does. Yeah. Love it.

When did you, when did you found Safe Software? Yeah, safe Softwares is company 30-year-old company that nobody, um, heard of. Um, both Dale and I were technical, so we just spent time, you know, building, um, you know, this solution and then selling it.

Um, and then, you know, much later we realized, you know, really we need to really get serious about our marketing. Um, and we sell around the world. Our first customer was actually in, in Sweden, if you can believe it.

And we're based in Canada. And now we have on the order of 20,000 customers around the world. Oh goodness.

From local governments to airports, to utilities. And again, any industry that cares about stuff that's on the ground is primarily, um, where we are embedded. Yeah, Love it.

Um, so another overnight success, you're telling me, huh? Yeah, That's right. 30 years.

I mean, and I tell people 30 years ago I had amazing hair and, and the funny Me too is Yeah. And the funny thing is older, the older I get, the better it was. Right?

Yeah, Exactly. And then I look at pictures from back then. Yeah.

You know, and it's like, you know, my hair was always kind of thin. I mean, not, not like now, but you know, it's never, I never had that like, you know, no craziness. But anyway, Hey, it happens to the best of us, Don.

Right? It does. It is what it is.

Um, so let's talk exp explicitly about what Safe Software does, as you said, and you gave us a little bit of background, but, you know, the world's changed, world's changing every day. Yeah. We're in this AI Yeah, Yeah.

Revolution. Yeah, That's right. As we sit here today, what's safe?

Yeah, yeah, yeah. So we are like, some people said, oh, we're ETL, but we did ETL many, many years ago. Now we just build like data virtualization, APIs, we have real time, high velocity, real time just connecting not just spatial systems, but all systems together.

I guess you could think of us like the nervous system of companies. So we have everything from the top companies in the Fortune, you know, 10 all the way down to the Fortune. Our market's like the Fortune 500,000, I guess you could say.

And, um, very big clients, very small clients. Um, but they all have one thing in common. They wanna get, you know, more value from their, you know, from their data.

So we basically chase data, you know, so OpenTable formats, for example, is a big thing. Um, AI and connecting data to AI is critical. 'cause AI needs data both for training, but also for inference.

And, and, um, we, you know, we do in more kinds of data than anybody. And, um, our whole model is any ai, because what is the best ai? It depends what you wanna solve, and it depends the data you have, right?

And the space is changing so quickly. So we want to give our customers that ability just to pick whatever ai, and we have some customers who are using three or four just to solve different problems, right? Yeah.

Absolute. I mean, no one wants to be locked in, right? Locked in's a really dirty word.

Locked is debt really more and more, right? So people want flexibility. So that a bit harsh.

Yep. Um, for people who, you know, want to engage with Safe Software, maybe wanna dig in deeper about what they're doing, how, how do they reach out to you, Don? Yeah.

com. Um, we're, we, we're technology people. We love talking to people about their data challenges.

We have a free 90 minute, you know, thing where you just show up and you're actually, we show you how to use the product. Um, it's a no code platform and, um, if you want to try it, we just give you the software. So again, and we'll help you get, you know, get value before you have to, uh, you know, um, learn about it.

We have webinars all the time, um, with customers. And so yeah. So just reach out happy to, you know, yeah.

To chat. Yeah. Very cool.

Um, Don, if you don't mind, I want to kind of pivot and talk about our topic of discussion today. Yep, yep, yep. Which is, uh, you know, that this whole debate about agentic AI is kind of irrelevant.

Um, what we're really seeing though is perhaps the extinction of what you call Fin SaaS mm-hmm. Mm-hmm. And vibe coding.

Yeah. Um, you know, and how this all fits in. Now, of course, you know, Satya Nadela, CEO of Microsoft.

Yeah. Yeah. He came out a few months ago and said, SaaS is dead or dying.

Yeah. Yeah. Uh, Killed by ai, right?

Mm-hmm. We're seeing it in the stock market. Big SaaS companies, service metrics, not service metrics, excuse me, service now.

Mm-hmm. Mm-hmm. You know, a monster of a company.

Their stock's down 50% this year. Yeah. And they, and they make their numbers.

It's not like they're not making their numbers. Salesforce, you know, kind of the granddaddy of, I don't know if I'd call that thin SaaS, it's Fat SaaS, but yeah. Salesforce one of the biggest SaaS companies down 35% this year.

Yeah. Yeah. How, what are you seeing?

How do you see it? Yeah, I mean, I, it's really interesting space because you, the, the, the key is you have to really have something. Having a database now with, um, you know, a, a thin UI on top of it isn't just isn't gonna cut it.

I was at an investor meeting, um, where, you know, companies came and, and this company came out, I believe it was Bain, and they talked about how when they're looking at a, a company to is to buy, they actually hire four people to see what, how much of that solution they can create in four months. And if they can create a, you know, the, you know, a significant portion of it, they, you know, then that tells 'em it's not, the company doesn't have anything, right? Because there's a lot of, there's a lot of companies out there now that have a very thin, very thin veneer that you can replace.

And for example, also at Safe, our sales team was gonna buy a tool. And I said to them, why don't you try to just recreate it with Gemini? And these are sales guys, so these are not technical people.

And they were able even, they were able to create it themselves to the point where they said, we're just gonna use Gemini. We're not gonna pay the, you know, the 80 grand a year for this other thing. Because in fact, it was this very thin veneer on top of some other LLM might've been Gemini, it might've been open ai.

You know, it doesn't, you know, it doesn't matter, right? And so, so, um, you need more than a good idea now. You need to really have something, you know, something substantial, some secret sauce that, um, you know, that really gives you extra value in the market.

Right. And, and, um, yeah, I, I agree with you. I, and I think you, you hit the nail right on the head.

What we're seeing is, you know, if I could recreate what you do mm-hmm. Using ai, what do I, I really don't need you. Yeah, that's right.

And we're even seeing it with open source software. Yeah. I have a lot of friends who, you know, big open source advocates, big open source consumers who are like, Hey, rather than dealing with the project and trying to get this functionality, you know, into the main branch or whatever, I'm, I'm just, I'm telling AI this is the functionality I need and it, it builds it for me.

Yeah. I don't need to use the open source software. Yeah.

Um, We're, we're seeing that as well, like our, our lead. And what's really interesting is our most experienced developers are the ones getting the biggest, um, the biggest accelerator than the the junior folks. Right.

Because again, they have that experience, they can see it. And so we're building an MCP server as part of our, our offering. And my lead developer, the guy's a genius, he came to me, he said this the, um, they're using Claude.

And he said, it's saved him weeks of work. Oh, Yeah. Right.

And of course, he's a very expensive resort. So to save him weeks of work is amazing. And so now I'm thinking, oh, you know, those weeks of work, it's not like he's gonna have weeks of not nothing to do.

Right. So it's, He's gonna do something else. It's gonna be able to creep.

Yeah. Yeah. Yeah.

So I, um, you know, I, I did a, i, every Thursday I do a thing called Shimmy, says I, yesterday Shimmy says, I, I spoke about this. Mm-hmm. We're creating a world where some people are gonna live below the AI poverty line.

Mm-hmm. Mm-hmm. And then some people are really going to excel and accelerate.

Yeah. Yeah. Because the way AI works, if you are a, let's call it a master coder, like a real great coder.

Yeah. The amount and quality of code you are gonna turn out compared to someone who is a novice junior and what they turn out with that same AI is gonna be just, you know, widely disparate. I mean, it's got, but it's not just code.

It's the same thing if you're a writer, you know, we produce content here. Yeah. It's the same thing if you're a writer, it's the same thing.

If you're a musician, I dunno if you've had a chance to say that, you know, synthetic music, they call it AI generated music. Well, in the hands of a real musician, it turns out crazy good stuff. Yeah.

You or I may not as good, you know, well, You, not Exactly the eye, the beholder, but, um, I think that's, you know, but it does beg the question though is, well, where do we get the next generation of great coders If the junior Guys, Yeah. I was an engineering company and this whole topic came up. They said, you know, they're the people with all the experience find they don't need the junior people anymore.

Um, and I was like, well, when they retire, who's gonna be, you know, so as a company, you need to hire your junior people and grow them up because you're, you're gonna need somebody to oversee the ai. And, you know, and they do stuff like Bridges. They're not going to take an AI generated bridge model and just, and Just let it roll.

Somebody has to Yeah. Somebody has to sign off on it and Yeah. You know, and when Bridges 97% accuracy is not good enough.

Right. You know, you need, you know, so anyway, really, so we say the AI is an assistant, not the authority. Right.

And I think others have said that, you know? Yeah. So I, so, you know, we've talked about it a lot around here at Techstrong and with people like you on Techstrong TV and some of our other shows.

Um, the way I look at it is fundamentally the job is changing. Yeah. And so if you were a computer coder and you spent a good portion of your day writing your code, testing your code, you know, squashing the bugs Yeah.

And, and all of that stuff, you're probably not going to do that. That's not going to be the definition of a coder tomorrow. Yeah.

No. The coder tomorrow is gonna be the guy who manages the AI or gal who manages the ai who writes the code, but who, number one could tell the AI exactly what I want out of that code. Right?

Yeah. What functionality, what characteristics. And number two has sort of, you know, how coders are when it's your own code, you can look at that code and you could see Yeah.

You know, you almost know it by memory. And if they, if something's outta place, it kind of sticks out, right? Yeah.

Yeah. You are gonna have to be good enough to look at that code that the AI generated. Yeah.

And see, oh, this looks, this looks good. Right. And you, I'm not saying you're not, you're not gonna test it.

Of course. We'll test it. Yeah.

But there might be another AI that does the testing, right? Yeah. That's right.

That's right. Yeah. And so you think about AI generated code, it has two code reviewers.

'cause it has the, the guy who's writing the code, he has to make sure he understands it. Yep. And then every line of code here, we also have another code reviewer.

Right. And because code review is critical. 'cause the most expensive part of, if code is not writing it, it's maintaining it.

And, you know, and we've had some pretty smart people here who could write really great code, but nobody could understand what it, how it worked. 'cause it was so complex. And so from a maintenance standpoint, it was just too complicated.

So we had to rewrite it in a simpler form that the mere mortal could understand it. Right. And so, you know, so it's gonna be interesting, you know, right now everybody's using AI to produce code, but we're gonna have to maintain it.

You can't, you know, so Yeah, I get it. I get it. Dawn, if it's okay, I wanna talk a little bit about vibe coding specifically, right?

Yeah, yeah. Um, it's fine. Who I was talking to a friend of mine the other day, very similar to your sales guy's experience.

Yeah. He, he, he just decided, you know what, I haven't coded in about the, the last time he coded, he told me it was punch cards. So that should give you an idea.

Should give you an idea. So he, but he, but he said, I, I gave it a shot. And you know what, in about an hour, you know, the hardest thing I had was not the code, it was just uploading the code to, to make the app run.

Yeah. And once I got that there and the code ran fine. Yeah.

He was amazed, amazed, amazed. Yeah. Yeah.

Um, You know, this vibe coding is gonna, we're gonna go from what do they think? There's maybe at most 50 million coders, 40 million coders in the world today. We'll have a billion coders or a half a billion tomorrow with this.

Yeah. And so again, AI with developers, Not coders. No, that's right.

But AI also hallucinates in code. Right. So it's gonna be interesting how, how high a qual I mean our, you know, our team uses Vibe coding.

Right. But then of course they, you know, and what we're doing right now is we have a no-code environment. We're using Vibe Code to create the No-code environment.

No Code. Yeah. Yeah.

So then you, uh, you know, so it's a, so we don't know what to call it. Is it a no-code vibe? No, it's Code Code.

But it's no Vibes. I don't know. No vibes.

The absence of vibes. It's, I, yeah. It's crazy.

This is the world. It's crazy. We live in though now.

Yeah. When you have milk bots finding religion and find founding churches and everything else that Goes like Yeah. Yeah.

There's whole social media that are only for bots, not for people. Right. Like Yeah.

And they have their own language. I know. It's crazy.

Well, No, now they wanna encrypt so that humans can't read, read what they're talking to each other about. So Yeah. You know, you, you didn't think of this 30 years ago when you started Safe Software.

No. And, and all of us had been working with AI for many years. Right.

And then all of a sudden chat, GPT Gen AI landed, and that really was the game changer. Right. We've been talking about machine learning and, you know, neural nets for years and we had libraries we had used, but nothing like this.

Right. This is, uh, yeah. This is one of those big moments.

And people talk about, is there an AI bubble? And I'm like, there was an internet bubble that burst, but the internet stayed Here, stopped The internet. The internet did not go away.

So, And AI's not going away either. It's not going away. There may be a financial reckoning, but that's another story's Right.

Technology's gonna stay. I mean Yeah, yeah. Absolutely.

But I'll, I'll give you another scary thing. com. Uh, so Claude released their latest Opus Claude Code.

Yeah. Or Claude Opus by Philanthropic. Mm-hmm.

6, like within a day or two of it being released, and the anthropic people setting it free, you know, setting it loose. It found like 600 vulnerabilities in a, across these open source projects. Yeah.

Across a whole bunch of like, important open source, not just obscure. Yeah. Yeah.

I don't know if we have the ability to deal with 600 new vulnerabilities every other, you know? Yeah. We just don't, we're not equipped yet.

We're gonna need AI to now respond to those Yes. Vulnerabilities And, You know, so it really is a, a new game. It's a new world.

Yeah. A company like Safe, right? Yeah.

You guys have been Yeah. Keeping current. And now That's right.

And we track vulnerabilities. And it's funny you mentioned that. 'cause one of the mandates for that team is to have, find AI that can find the vulnerabilities in all the code that's being written.

Right. And, and also the testing and, you know. Absolutely.

Yeah. Yeah. It's a great time.

I mean, look, of all the timelines you could have been born in, Ah, it's, we were lucky. I mean, really the PC was just powerful enough that two crazy technologists could start this company. And then, and then the importance of data, you know, for the last 30 years is only getting more and more important all the time.

And now we're running outta data for ai. So we talk about synthetic data. Yeah.

You know, creating our own data. I don't know if that's like synthetic oil for the car or not, but, uh, but you know, you get the idea, right? Like Yeah.

Oh, I get it. We, I mean, we're all looking at it. And, and I think, you know, the irony of ironies is that we use AI to police ai, to, we use AI to test ai.

Yes. Um, You know, it, it's, it's a great time to be in here, is all I got to say, Don, we're about outta time. One more time.

Okay. What was that? com.

com. Yep. Perfect.

Yep. Awesome. Yeah, I wish you success for the next 30 years, but don't wait that long to come back here on Text Strong tv.

Okay. No, It's been, it's been a real pleasure and, uh, great talking with you, Alan. Yeah.

Really good. All right. Okay.

Don Murray, co-founder and CEO of Safe Software here on Textron tv. We're gonna take a break. We'll be back.