Founding Thesis · 2026-04-14
Operator-led businesses have been priced out of real software for decades. AI just changed the math.
The labor cost of building real software has collapsed. Not in theory. In practice. A capable engineer with good AI tooling can now ship alone, in weeks, what used to take a four-person team a quarter. The economics of who can afford to have software built — and what kind of software — have quietly flipped over the last two years.
That shift matters most for the businesses that have been locked out of custom software for decades. Businesses doing $2M to $50M in revenue. Operator-led, hands-on, often excellent at what they do. Running on spreadsheets and email for work that should live in software, patched over with human effort because the alternatives were economically out of reach.
These businesses know they have a problem. They've tried SaaS, and most of it solves part of what they need while creating three new problems they didn't have before. They've looked at custom development and decided, correctly, that it doesn't pencil out.
So the work keeps living in spreadsheets. And now, suddenly, the math is different. But almost nobody has told them, and almost nobody is set up to do the work.
I left my engineering role to be one of the people who does.
If you're running one of these businesses, there's a decent chance you've stopped talking about this problem out loud. You know exactly which workflow is eating your team's week. You've probably thought about building something, or hired a consultant who produced a deck, or tried a SaaS tool that didn't quite work. At some point you filed the problem under "this is just how it is" and went back to running the business. What I want to argue in this post is that the ground has moved under that decision, and it's worth reopening.
The easy story is that these businesses are behind the times. That they need to "embrace technology," or "digitally transform," or whatever the consultant word is this year. That's not what's happening. If you're running one of these businesses, you know exactly what software could do for you. You've been pitched SaaS by every vendor that'll take your meetings. You've tried things. You've paid for tools you don't use.
The real story is harder, and more interesting.
Off-the-shelf SaaS doesn't fit because the work is specific. A generic CRM doesn't understand how your hotel group handles group RFPs between five properties, or how your e-commerce brand manages wholesale reorders across dozens of accounts, or how your property management company tracks open maintenance requests across a portfolio. The software that does exist is built for the average of ten thousand businesses, and every actual business is the specific case, not the average.
What usually happens is worse than no software. A generic tool gets adopted, it solves forty percent of what the team needs, and the other sixty percent gets stuffed into spreadsheets that live alongside the tool. Now there are two systems of record, the official one and the real one, and the team spends time reconciling between them forever. This is the actual state of most mid-size businesses today. Not "no software." Partial software, with all the operational drag that comes with maintaining the gap.
Custom-built software would fit. But until recently, custom software has been economically out of reach for any business that wasn't either enormous or venture-backed. A capable in-house engineering team runs $800K+ a year in salaries alone. A good agency charges six figures for an engagement and then hands you something you have to maintain yourself. For a $10M company running on 8% margins, neither option makes sense.
So the work keeps living in spreadsheets. And eventually, the people running the business stop bringing it up, not because the problem went away, but because there was no version of solving it that made sense. This is the quiet part. If I had to guess, most operators can describe their single biggest operational pain in one or two sentences. What they've stopped doing is expecting anyone to fix it.
The businesses I'm describing know something has changed. They've read the articles. They've sat through the board meetings where someone says "we need an AI strategy." They've used AI tools themselves and seen them work. What they don't have is a version that's repeatable, owned by the business, and doesn't depend on one person's browser tab.
What they do next is usually one of two things, and both are wrong.
Some of them buy AI features from their existing vendors. If the underlying tool didn't fit your business before, turning on the "AI" checkbox doesn't help. You get a more capable version of something that was already solving the wrong problem.
Some of them try to build or automate it themselves — an employee spins something up with AI coding tools, or someone spends weeks configuring Zapier flows and no-code automations. The opportunity cost is real: that's operator attention pulled out of running the business. And what gets built tends to be fragile — nobody fully owns it, it breaks at inconvenient moments, and the security and maintenance burden quietly accumulates in the background.
What almost none of them do is the thing that actually works: find someone who can sit with the business, understand the specific workflow that's costing them the most, and build real software around it. Not a bolt-on feature. Not a fragile automation that breaks when the workflow changes. A real application, owned end-to-end, that fits how the business actually operates.
The reason almost no one does this is because, until recently, it was too expensive. Now it isn't. But the market hasn't caught up to that fact yet. There's a gap between what's now economically possible and what most businesses know to ask for. That gap is where the interesting work is.
Before starting Cymba Labs, I was building production AI systems used inside hospitals. Real systems, with real stakes, integrated into workflows where being wrong has consequences. That work taught me two things that matter for what I'm doing now.
The first is that the rate at which I could ship meaningful work accelerated past anything I thought was possible. That's the shift I was describing at the top of this post. I've lived the economics of it, not just read about it.
The second is that building software that holds up under real operational load is a completely different skill than wiring together a demo. Working in healthcare, where edge cases have consequences, where integrations have to survive audit requirements, where someone has to figure out why the system got something wrong, teaches you the difference between software that works and software that looks like it works. That experience is not glamorous, but it's what separates a system you can trust from one you have to babysit.
Both of those things together, shipping fast with modern tooling and knowing how to build things that actually hold up, are exactly what operator-led businesses need. And until recently, you could only get both from a company with a real engineering department. That's the gap this shift has opened up.
I started Cymba Labs to be one of the studios doing this work.
There are millions of businesses in the United States sitting in the range I described, operator-led, excellent at what they do, running on patched-together software because the real thing was never affordable. Independent hotel groups running rate strategy in Excel. E-commerce brands managing wholesale reorders over email. Property managers fielding maintenance requests over text and tracking them in a shared spreadsheet. The cost of that patching, in wasted time and operational drag, is enormous in aggregate but hard to see from the inside because it's just how the work has always been done.
What makes this more than just cheaper software is that the same AI that's changed how I build can also be embedded in what I build. The businesses I'm describing don't just need their spreadsheet workflows replaced with a form. They need software that can respond to a booking inquiry, process a wholesale reorder, or dispatch a contractor — without someone sitting at a computer to make it happen. That's the actual shift: not just that custom software is now affordable, but that affordable custom software can now do things that used to require a person.
I work with a focused number of operator-led businesses at a time. I'm currently building across hospitality, e-commerce, property management, and professional services, among others.
If any of this resonates — if you've been quietly patching over a workflow you gave up on solving, or if you're trying to figure out what to do with AI in your business and haven't gotten a straight answer from anyone — I'd like to hear about it. No pitch. Just an honest conversation, and I'll tell you whether I think I can help.
Reach me at charlie@cymbalabs.com.