The AI bill nobody budgeted for

Two of the most sophisticated tech companies on earth just hit a wall with AI. And not the kind you would expect.

Microsoft is pulling most of its Claude Code licenses across the division that builds Windows, Outlook, and Teams. Uber burned through its entire 2026 AI budget in four months. Gone before the year was halfway over.

Not because the tools failed. Because they worked too well.

If two companies with effectively unlimited budgets cannot keep their AI spend under control, it is worth asking a quieter question: what happens to the rest of us?

The Part Most People Miss

Most AI tools are sold on usage. You are shown the features and the demos, but what you are actually buying is a meter. Every prompt, every project, every new idea your team runs through it adds to the bill.

It is a brilliant business model for the vendor. The more your team leans in, the more they pay. The more comfortable your people get, the faster the number climbs.

Now flip it to your side of the table. As your team gets good at AI, more prompts, more builders, more projects, your cost does not level off. It compounds. The success you wanted quietly becomes the expense you never planned for.

The Numbers Tell the Story

At Microsoft, roughly 5,000 engineers adopted Claude Code. Adoption climbed past 90 percent. Per-engineer costs ran $500 to $2,000 a month. The tool was not cut because it was bad. It was cut because, in their own words, it had become too popular.

Uber tells the same story from the other side. Its engineers love the tools. About 95 percent use them, and roughly 70 percent of new code now comes from them. And yet Uber's own COO cannot connect that spend to results. His words: "That link is not there yet."

Loved by the team. Impossible to budget. No clear return. That is the trap in one sentence.

You Are Not Uber. But the Structure Is the Same.

You are not spending billions on engineers. But the trap was never about size. It is about structure.

If your AI is billed by usage, your most engaged month is also your most expensive month. For a tech giant, that is a headline. For a mission-driven organization on a fixed annual budget, it is something more serious. An invoice you cannot predict, attached to a tool your team has come to depend on. That is not a line item. That is a risk to the mission.

Why We Built FRANSiS the Way We Did

I think about this a lot from the operator seat. Because I have watched what happens when a team falls in love with a tool and then gets the bill.

We made a deliberate decision early on. FRANSiS partners pay one flat monthly rate for unlimited use. No credits. No per-message pricing. No meter running in the background. We carry the variable cost so you never see it.

That changes the whole relationship. Most vendors quietly profit when you use less. We only win when you use more, because the more of your communication FRANSiS handles, the more value you get for the same predictable rate. Our job is not to ration your usage. It is to help you push the platform as far as it can go.

You commit to a number you can plan a whole year around. Then we go to work making that number return as much as it possibly can.

The Question Worth Asking Before You Sign Anything

Before I would sign with any AI vendor, I want one thing answered: does this help me scale my mission and my impact, or just my monthly invoices?

Microsoft and Uber learned which question matters. They just learned it after the fact.

You do not have to.

‍