Practitioner essays on building production software with AI agents.

This is the third time your company has tried to fix the monolith. The first two attempts cost millions and changed nothing. AI agents make the third attempt possible, but only if the person leading it has actually done this before. Here is how to tell in ninety seconds.

Your engineering team reports a 30% velocity improvement from AI tooling. Your CFO sees 5% on the P&L. The gap, millions per year, is sitting in your value stream, visible to anyone who picks up a marker. A whiteboard and a Tuesday afternoon will show you exactly where the money goes.

Every leader I ask the fifty million dollar question gives the same answer. They would not build what they have now. They would start over from first principles with something smaller, leaner, and faster. That answer tells me whether they understand how software gets made, or whether they are just running the theater.

Your product team spends 60% of their time on coordination that agents handle in minutes. Building is no longer the bottleneck, your customer's capacity to receive what you ship is. The operating model that wins is embarrassingly simple: talk to customers, figure out what to build, ship it, make sure it lands.

Your revenue team already exports pipeline data out of the CRM to do the real work somewhere else. You are not buying a strategic platform. You are paying $2.2 million a year for a database you already own, and four engineers could replace it in twelve weeks.

Your CFO should absolutely control the money. Your procurement team should absolutely manage the buying process. But neither of them should be deciding which AI capabilities your company adopts, because AI is not a printer.