For most of the last twenty years, custom software was a luxury good.

If you ran a small business — a dental office, a tile shop, a two-truck plumbing company — you had two options. Pay an agency $60,000+ for a bespoke tool, or squeeze your workflow into whatever generic SaaS happened to exist. Most people picked option two and quietly accepted that the software would never quite fit.

AI is breaking that tradeoff. Not loudly, not overnight, but in a way that’s going to reshape who gets to have software built for them.

Why custom software was always out of reach

The cost of custom software was never really about the code. It was about everything around the code.

A small business owner needed a working CRM with three custom fields and an SMS reminder. To get that, they had to pay for:

  • An agency to scope the project
  • A designer to mock up screens
  • A backend developer for the API
  • A frontend developer for the UI
  • A QA pass before launch
  • An ongoing retainer for bug fixes and small changes

The actual feature set was maybe two weeks of real work. But the coordination, billing overhead, and risk pricing turned a simple tool into a six-figure project. So small businesses didn’t get custom tools. They got Salesforce. Or a spreadsheet.

What AI actually changes

The honest framing isn’t “AI writes the code now.” It’s that AI collapses the team.

A solo developer with Claude Code or Cursor can now do what used to require four or five people. Not because the AI is brilliant — it isn’t, not yet — but because it absorbs the parts of the job that used to require specialists. Boilerplate, glue code, basic UI, schema migrations, deployment scripts, integrations with Stripe or Twilio. The grunt work that used to eat 70% of an agency’s hours is now a prompt away.

What’s left is the part that actually matters: understanding the business, choosing the right shape for the tool, and making sure it works end to end. A single experienced developer can hold all of that in their head — and ship the whole thing in a few weeks.

That changes the math completely. A custom tool that used to cost $60,000 because it had to fund a five-person team for two months can now cost $8,000 because it funds one person for two weeks.

The shape of the new market

I think we’re at the start of a quiet shift in who builds for small businesses.

The traditional dev shop model — billable hours, account managers, scope creep — doesn’t survive this. The economics only worked when you could justify a $200/hour blended rate against a team. When one person plus AI can deliver the same outcome, the agency’s overhead becomes dead weight.

What replaces it looks more like a freelancer or a tiny studio. Two or three people, deep familiarity with one industry, AI doing the heavy lifting. They charge a flat fee that a local business can actually pay. Five thousand dollars for a custom intake form that beats every SaaS option. Three thousand for an internal dashboard that took a week to build. That’s a price point small businesses understand — it’s roughly what they’d spend on a piece of equipment.

And unlike SaaS, they own the result. No per-seat fees. No vendor going out of business. No “we’re sunsetting that feature next quarter.”

What small businesses actually want

The mistake a lot of builders make is assuming small businesses want what big companies want. They don’t.

A dental office doesn’t want a dashboard with twelve KPIs. They want a tool that texts patients the day before their appointment and stops texting them once they reply “yes.” That’s the whole feature. A tile shop doesn’t want enterprise inventory software. They want a way to mark slabs as reserved when a contractor calls so they don’t double-sell.

These are tiny, sharply-scoped tools. They’re embarrassingly simple from a software standpoint. And they were impossible to justify building under the old cost structure — no agency was going to take a $4,000 project, and no SaaS was going to build something that only applies to one type of shop.

Now they can exist. That’s the real unlock.

The risk worth naming

I don’t want to oversell this. AI-built software has the same failure modes any rushed software has — and a few new ones.

If a developer leans too hard on the AI without understanding the output, the small business ends up with a tool that works on day one and breaks subtly on day ninety. Edge cases nobody thought about. A Stripe webhook that silently stops firing. A timezone bug that miscounts appointments. The cost of fixing those after the fact can erase the savings from building cheaply.

The right move for small businesses isn’t “hire the cheapest person with AI.” It’s hire someone who actually understands the code they’re shipping, even if they used AI to write most of it. The price point dropped because the team got smaller — not because the rigor did.

Where this lands

For the first time, the price of custom software is in reach for the kind of business that’s never been able to afford it. A bakery, a landscaper, a dance studio — these are companies that can fund a small bespoke tool now, and they couldn’t five years ago.

That doesn’t mean every small business needs custom software. Most will keep using SaaS, and that’s fine. But the businesses that have a genuinely unusual workflow — the ones who’ve been losing hours every week to a tool that almost fits — finally have a way out.

The agencies built the old market. AI is building the new one. And the people who win are the small businesses who never used to make the cut.