Last year, building a custom internal tool took weeks of developer time and a strong argument for why you couldn't just buy something off the shelf. Now, with AI coding assistants, you can scaffold that same tool in an afternoon. But here's the catch: maintaining it still costs exactly what it always did.
This is the core argument in a recent analysis from Runframe, and it deserves attention because it reframes a decision every team is making right now.
Building Is Cheap. Owning Is Not.
AI coding tools like Cursor, Claude Code, and Aider have genuinely compressed the build phase. A solo developer can produce working software faster than ever. The temptation to build custom solutions has never been higher, because the upfront cost has cratered.
But software isn't a one-time expense. It's an ongoing liability. Every custom tool you build needs bug fixes, security patches, dependency updates, feature requests from teammates, and someone who understands how it works when the original builder leaves. None of that got cheaper. AI can help you write a patch, sure, but someone still needs to notice the bug, diagnose it, test the fix, and deploy it.
The maintenance burden scales linearly with the number of custom tools you operate. Build five internal tools in a week because AI made it easy? Congratulations, you now maintain five internal tools forever.
The New Build-vs-Buy Math
The old calculus was simple: building is expensive, so buy unless you have a very specific need. The new calculus is trickier. Building is cheap, so teams are building more. But the total cost of ownership hasn't changed, it just shifted. You're paying less upfront and more over time.
This means the right question is no longer "can we build this?" (the answer is almost always yes now) but "do we want to own this?" Owning means on-call rotations, documentation, onboarding new team members, and handling edge cases that AI-generated code didn't anticipate.
For something like incident management (the specific example in Runframe's analysis), the buy case is stronger than ever. The build cost dropped, but the operational cost of running your own incident management system - integrations, alerting rules, post-mortem workflows, compliance requirements - didn't budge.
Where This Leaves Teams
The practical takeaway: use AI to build things that are truly unique to your business. The stuff no vendor sells. But for commodity infrastructure - project management, incident response, CRM, monitoring - the maintenance tax makes buying the obvious choice, even when building feels trivially easy.
The most expensive software isn't the hardest to build. It's the easiest to build but hardest to stop maintaining.