I’ve spent twenty-five years running sales and marketing teams, and along the way I’ve hired a lot of developers. Good ones, mostly. And almost every project had the same friction point, which I can describe in one sentence: I speak business, they speak code.
That’s not an insult to engineers. It’s a description of how the work was divided. Engineers are trained to build things. They are not trained to run a quarter, sit in a pipeline review, or explain to a founder why the ad budget moved. So when you hand an engineer a problem, you first have to translate it out of the language your business actually runs in. Then they build something, and you translate back, and you find out what got lost. Every round trip costs weeks and money, and the people billing for those weeks have no reason to shorten them.
For decades, that was just the price of custom software. The people who knew what to build couldn’t build it, and the people who could build it didn’t know what to build. Since writing code was the scarce skill, the leverage sat with the builders. If you couldn’t afford them, you got the consolation prize: SaaS.
And look, SaaS was a genuine revolution. It’s also a compromise we’ve all stopped noticing. Every tool in your stack was designed to work for the widest possible set of businesses, which means it was designed for none of them in particular. Software that works for everyone is perfect for no one. That’s why your team is juggling logins, exporting CSVs, and running a business through the gaps between six products that don’t talk to each other.
Here’s what changed: AI collapsed the cost of writing code. Not by a little. Building custom software got ten to fifty times cheaper in about two years. Which means writing code is no longer the scarce skill.
Knowing what to build is.
Think about what that inverts. The judgment that used to be the cheap part of the equation, the part you couldn’t bill hours for, is now the whole game. Knowing that your reps will abandon any tool that adds a step to their day. Knowing which report the founder actually reads. Knowing that “integrate with our CRM” means one specific workflow and not forty features. That knowledge only comes from having run the business, carried the number, and lived with the tools. You can’t prompt your way to it.
So the leverage moved. It moved from the people who can write code to the people who know what the code is for. Operators.
Two honest caveats, because this argument gets oversold. First, operator-led doesn’t mean amateur-built. AI writes the code faster than humans do; it doesn’t decide what ships. Anything worth running a business on still gets tested, secured, and maintained to engineering standards. The advantage isn’t skipping the discipline. It’s skipping the translation layer.
Second, most of the industry has noticed this cost collapse, and their plan is to keep charging you the old price and pocket the difference as margin. The build got 10x cheaper; the invoice didn’t. Watch for that.
What I’d tell any founder-led business right now: the software you actually wanted, the one that fits how your company really works, stopped being an enterprise luxury. The constraint was never your budget. It was the cost of translating your judgment into code, and that cost just fell through the floor. The people best positioned to take advantage of it aren’t in engineering departments.
They’re the ones who already know exactly what needs to be built. They’ve known for years.