AI Made Building Software Easy. That Is Exactly Why You Should Think Twice Before You Build.

A founder told me last month that his team had rebuilt a piece of software over a weekend that they had been paying a vendor $40,000 a year for. He was thrilled. I asked him a single question: who owns it at 2am when it breaks during a product launch? The room went quiet.

This is the new shape of the build versus buy decision, and almost everyone is getting it wrong in the same way.

The old calculus is genuinely broken

For twenty years the logic was simple. Building software was slow and expensive, so you built only the things that were your core differentiation and you bought everything else. Payroll, support tooling, analytics, CRM, you rented all of it, because writing and maintaining that code cost more than the subscription ever would.

AI broke the first half of that equation. A capable engineer with a coding assistant can now stand up a working version of most internal tools in days, sometimes hours. The prototype that used to justify a six month vendor evaluation now appears before the evaluation meeting is even scheduled. So the instinct is understandable: if we can just build it, why are we paying for it?

Because the cost of software was never the building. It was everything after.

The weekend is the cheap 10 percent

The demo that works on Friday is the part AI made cheap. It is also the part that was never expensive. The expensive parts are still exactly as expensive as they always were, and AI did almost nothing to touch them:

• The edge cases you have not thought of yet, which show up as angry customers, not as compiler errors.
• Security, permissions, audit trails, and the compliance questions that surface the first time a real customer's data flows through it.
• The integrations that break every time an upstream system changes, silently, at the worst possible moment.
• The person who has to be on call for it, forever, instead of the vendor whose entire company is on call for it.

A vendor is not selling you code. They are selling you a team of people whose full time job is to keep that code working while your team sleeps. When you rebuild their product over a weekend, you have bought the code and quietly signed up for the team, except the team is now one of your own engineers who also has a day job.

AI collapsed the cost of the prototype. It did nothing to the cost of ownership. Confusing the two is how a $40,000 subscription becomes a $200,000 distraction.

The question that actually matters now

The old question was can we build this. AI made the answer almost always yes, which means it is no longer a useful question. The new question is sharper: is this thing a source of our advantage, or a cost of doing business?

If it is a source of advantage, something your customers feel, something a competitor copying your vendor could not replicate, then build it, and AI now makes that far more affordable than it was two years ago. This is the real gift of the moment. Things that were too expensive to build custom, and so everyone rented the same generic tool, can now be tailored to how your business actually works.

If it is a cost of doing business, something every company in your industry needs and none of them win on, then buying is still almost always right, and AI did not change that. Nobody has ever won a market because their internal expense reporting tool was bespoke.

A decision rule you can use on Monday

Before you greenlight the weekend rebuild, force three answers onto a single page:

• Differentiation. If this worked perfectly, would a customer ever notice or care? If the honest answer is no, you are about to build a liability, not an asset.
• Ownership. Name the person who maintains this in eighteen months, after the engineer who built it has moved teams. If you cannot name them, you have not finished the decision.
• Total cost. Compare the subscription not to the build cost, but to the build cost plus a realistic estimate of maintenance, on call, and the opportunity cost of that engineer not working on your actual product. That last number is usually the biggest and always the one left off the spreadsheet.

Run those three and the weekend rebuild that felt obvious often stops feeling obvious. Sometimes it survives, and when it does, you build with real conviction instead of the thrill of a working demo. That is a much better place to start.

The uncomfortable summary

AI did not make the build versus buy decision easier. It made it easier to build, which is a different thing, and it made the wrong choice cheaper to start and just as expensive to live with. The teams that will win the next few years are not the ones that build the most. They are the ones that stay ruthless about what is worth owning.

If you want the full framework we use to run this call with clients, including how we score total cost of ownership, we wrote it up here: Build vs Buy AI: a Decision Framework for Mid-Market Teams. And if you are staring at a build versus buy decision right now and want a second set of eyes, that is a good part of what 10dem does.