Automation and money: why successful pilots blow up a company's budget

You know that feeling when you sign up for a free trial of a service, everything works great, and then the annual bill for the whole team arrives and the CFO's eye starts to twitch? That is roughly what is happening now with enterprise automation and AI adoption.

Greg Holmes, CTO of Apptio (which IBM acquired not long ago), has decided to play the friend who spoils the party by asking: "So who pays for all this?" The core of his message is simple but painful: the "Field of Dreams" model of technology adoption, build it and they will come, is officially dead. At least when it comes to intelligent automation.

What exactly is the problem? The "pilot trap." Launching a small bot or script in one department is cheap and fun. You show pretty slides, efficiency goes up, everyone is happy. But as soon as you try to stretch that approach across the whole corporation, the math breaks. Infrastructure costs, licenses, cloud computing and support start eating the budget faster than the automation delivers value.

Apptio is not merely theorizing here. After IBM bought it for $4.6 billion, it ended up with a huge body of data on how companies spend money on IT. And it sees a pattern: executives often confuse the technical ability to scale with financial viability. What worked on a developer's server or within a single department requires a completely different level of financial rigour at the enterprise level.

This is a classic FinOps problem, and the arrival of AI and GenAI has only made it worse. We used to forget to switch off virtual machines; now we forget to optimize tokens and agent requests. Holmes insists that without a transparent link between automation costs and concrete business value, scaling is just a way to burn investors' money faster.

The bottom line: the era of limitless "digital transformation" budgets is over. If you cannot show the CFO exactly how your new AI agent pays back every cent spent on it in the cloud, the project will most likely be axed before it ever leaves the MVP stage.