How Jamf Stopped AI Costs From Quietly Exploding
AI can secretly rack up huge bills at work. Here's how one company stopped it.
Here's the problem that many companies are starting to hit: AI tools don't cost money like regular software. With old software, you pay for a set amount of computing power ahead of time. With AI, you pay based on how much people use it. And because AI usage is invisible in the moment, the bill can arrive as a nasty surprise. One engineer testing an AI coding assistant for a few hours can burn through more budget than an entire team spends in a week.
Jamf, a company that manages Apple devices for businesses, gave all of its engineers access to AI coding tools through Amazon Bedrock (a service that lets companies use AI models like those from Anthropic). Productivity went up, but so did worry about the money. So Jamf built what tech people call a 'tokenomics' system — but in plain English, it's a financial safety gauge for AI usage.
The solution works automatically. It tracks how much each engineer spends each day on AI, right down to the person. When someone reaches 80% of their daily budget, the system quietly locks away the “premium” AI models (the most powerful and expensive ones). When they hit 100%, it blocks the mid-tier models too. But it leaves the cheapest model available, so engineers can keep working without wasting time. The restrictions come into effect within minutes, and they disappear the next day, so nobody gets stuck permanently.
Why should you care? If your company uses AI — or plans to — this kind of control is what makes it affordable. Without it, companies either get shocked by bills or simply say no to AI. Jamf's system proves there's a middle road: use powerful AI freely, but with guardrails that protect budgets. For everyday workers, it means AI tools can keep spreading through your workplace without your boss pulling the plug after the first expensive month.
- AI costs grow based on how much people use it — invisible until the bill arrives.
- Jamf's automated system tracks each employee's daily AI spend and blocks expensive models when limits are hit.
- Engineers still have a cheap AI model available, so nobody's work stops — just the expensive parts.
Why It Matters
If AI costs can't be controlled, companies will stop offering them. This shows a way to keep AI both powerful and affordable.