TL;DR: Your engineering team bought a €200/month AI tool. Then marketing bought one. Then sales bought three. By the time anyone notices, you're spending six figures on tools nobody centrally manages. I've watched this happen at six companies now. Here's what it looks like, why it keeps happening, and what to do about it.
It never starts with a big decision. It's always small. A team lead signs up for something to try. It works well enough that the team keeps using it. They expense it. Nobody questions a €50 charge.
Then another team does the same thing with a different tool. Then another.
Six months later, you have fourteen AI subscriptions across four departments. Nobody can tell you what each one does or whether they overlap. The total is higher than your cloud bill.
Three of them were different AI writing assistants. Two were variations of the same code completion plugin. One had been running for eight months with no activity.
Traditional procurement doesn't work for AI tools. They're too cheap to trigger approval workflows. There's no IT gate because most of them are SaaS products you can buy with a credit card. And every department head has a budget line that says "tools" with no further breakdown.
The incentives are also backwards. Individual teams are rewarded for moving fast. Buying a tool is faster than waiting for central IT to approve something. So they buy. Nobody is rewarded for saying "let me check if we already have something that does this."
Don't try to block shadow AI. You'll fail — people will just expense things differently. Instead, do this:
1. Pull the last three months of expenses tagged "software" or "tools." Categorize everything that looks like AI. You'll probably find 15-30 subscriptions. Most of them will be duplicates.
2. For each one, figure out who's using it and for what. Most teams don't know what other teams are using. Just showing them the list often triggers consolidation on its own.
3. Make it internal. "Here are all the AI tools we're paying for as a company." If two teams see they're paying for the same thing, they'll merge naturally.
4. Not a procurement process. One rule: before buying a new AI tool, check the list. If something already exists, use it first. That's it.
Teams that do this typically cut 30-40% of their AI spend in the first month. Not because they stop using AI. Because they stop paying for duplicates and abandoned tools.
The real win isn't the cost savings. It's that you now know what AI your company is actually running. You can't govern what you can't see. Once you can see it, the hard part is over.
*This is the kind of thing I write about at [Ataraxium](https://ataraxium.ai). We build tools for AI operations, cost governance, and compliance.*