The multibillion-dollar custom software industry just became a commodity anyone can spin up over lunch.
The Summary
- Paul Ford, decades-long software veteran, says AI has fundamentally changed who can build software by collapsing the cost of custom enterprise tools.
- "Vibe coding" and AI code generation promise to turn anyone into a software company, but Ford warns this doesn't mean everyone becomes a coder.
- The real shift: custom software tailored to individual needs is now economically viable where it was previously reserved for enterprise budgets.
The Signal
Most software in the world isn't the apps on your phone. It's the boring, expensive, bespoke tools companies pay armies of developers to build for managing sales, inventory, workflows. Paul Ford, who runs New York-based business software agency Aboard, calls this the "custom and enterprise" bucket. It's where the real money flows and where AI code generation just turned the economics upside down.
What Ford describes isn't theoretical. AI tools can now write functional code from plain language descriptions. The barrier between "I need software that does X" and "here's software that does X" has collapsed from months and six figures to hours and the cost of a ChatGPT subscription.
"Something that was very expensive is now available to everybody."
But here's where Ford parts ways with the "everyone will code" crowd. He doesn't think we're heading into a world where your accountant is debugging Python at 11pm. The future he sees is weirder and more useful: custom software becomes so cheap to generate that everyone gets tools built exactly for their workflow, not the other way around.
Think about what enterprise software costs now:
- Custom CRM buildouts: $100K minimum
- Industry-specific workflow tools: mid-six figures
- Time from spec to deployment: 6-18 months
AI code generation compresses all three. Ford's betting this doesn't just make software cheaper. It makes personalized software viable at individual scale. Your weird use case that would never justify a dev team? Now it's a Tuesday afternoon project for Claude or whatever comes next.
The Implication
If Ford's right, the software industry splits. One tier keeps building the foundational models and platforms. The other tier becomes a sea of hyper-specific, AI-generated tools that wouldn't exist under the old cost structure. The middle tier of custom dev shops building standard enterprise CRUD apps gets squeezed hard.
For everyone else, watch for software to start feeling less like "here's what the vendor built" and more like "here's what I needed today." The test isn't whether you learn to code. It's whether you can describe what you want clearly enough for an AI to build it. That's a different skill, and it's about to matter a lot more than syntax.