Since 6 July, Tesla has limited spending on third-party AI tools to $200 a week per employee. The company joins Uber, Meta, AT&T and Amazon in the same wave of restrictions: a sign that the bill for autonomous AI tools is becoming a governance issue in its own right.
The facts
- Uber had used up its annual AI budget in four months before capping its coding tools at $1,500 a month.
- The heaviest-spending companies reportedly spend up to $7,500 per employee per month on AI tools.
- Amazon scrapped its internal AI usage ranking after employees gamed it.
Why do AI budgets slip out of control?
The mechanism is simple: recent AI tools are billed on usage, each team subscribes on its own, and nobody consolidates. When Uber uses up its annual budget in four months, the tools are clearly being used; what is missing is a framework linking the spending to a measured use. The caps decided by Tesla, Meta or AT&T are the emergency response to a visibility problem: knowing who uses what, for what gain.
The episode of Amazon's internal ranking, scrapped after employees gamed it, illustrates the other pitfall: measuring AI consumption instead of its value encourages the wrong behaviour. The right metric is the time saved, on which tasks, at what cost, well ahead of the number of queries. This logic holds at every scale: an SME that piles up AI subscriptions with no inventory or tracking reproduces, on a smaller scale, exactly the drift these giants are now correcting.
AIxH's view
Giving access to AI tools with no tracking framework always ends up costing more than planned. That is precisely what an AI maturity audit covers: taking stock of the tools actually used, knowing where the budget goes, and reallocating it to the uses that produce a measurable gain.