Your AI Chargeback Tells You Who Pays. It Doesn't Tell You What Worked.

November 09, 2026

Watch the video version:

Late in October, I sat at a roundtable in New York with technology executives from organizations far larger than most of the teams I work with. The conversation kept circling one question: who pays for AI? One executive described how their organization settles it. Every business unit pays the cost of its AI tools back to a central data and AI office. If someone has not touched a tool in ninety days, nobody asks why. The access is removed.

A second leader described the same move from the other side. Their AI spend still sits centrally, inside the CIO's budget, and the fight in their 2027 budget planning is to push it out to the business units as usage and token costs climb. Their reason was blunt: with the cost held in the center, it looks like the company is doing AI, and it is not really doing it. A third leader cut through both with one question: "What are you getting for the dollars you're spending?" These are their accounts of their own organizations, and I have not checked their numbers.

The standard answer is accountability. Stop funding AI from a central budget, charge each unit for what it uses, and pull the licences nobody touches. Put the cost where the usage is, and spending turns disciplined. If you are writing a 2027 budget, you have heard some version of this in planning meetings this fall. Chargebacks get presented as what a mature AI program looks like, and underneath sits an assumption nobody says out loud: whoever pays will measure value.

Run the first model forward. The unit pays, and the central office decides who keeps access. The bill moved; the decision stayed where it was. The people who own the cost are not the people who judge the result, and the only judgment in place is a usage count. Anyone who wants to keep a tool logs in often enough to keep it, and the dashboard reads every one of those logins as adoption. Usage was the scoreboard while the cost sat in the center, and the ninety-day cutoff keeps it the scoreboard, now with a penalty attached.

Then the unit does what anyone does with a budget line it pays for and cannot steer: it protects the line. The tools with visible usage stay, because usage is what the center counts. The trials with slow or uncertain payoff go first, and those are the bets that may have found the value. Nobody wrote down who decides whether an AI bet worked, on what evidence, or by when, so that question defaults upward and lands back on the CIO's desk at budget season. The budget cycle makes the call, the same way a hiring freeze can end a role nobody actually decided to end, and the organization ends up with spending that looks disciplined and one learning loop a year.

If you are the one moving the bill, you are the one who splits it from the decision, and you are also the one person who can keep them together. The business unit I lead IT for runs almost entirely on its own. Every IT cost it has lands against our unit, and that is the right way to do it. It works because the leadership that pays for a tool is the same leadership that decides whether it earned its cost, on results we can point to. Cost and authority sit in the same place.

The organizations that make chargebacks work do not have better dashboards than yours. They moved the decision with the bill. That is what I call Decision Boundaries: the explicit definition of who owns which decisions, under what conditions, and when to escalate. For AI spend, the boundary fits in one question, answered in writing before the bill moves: who can end this AI bet, and on what result? If the answer is the unit that pays, move the bill. If it is anyone else, the bill is not ready to move.

Take the AI tool your team spends the most on. Name the one person who could cancel it tomorrow, and the result they would point to when they did. If you named a login count, or someone outside the team that pays for it, the bill and the decision have already split. You can close that gap before your 2027 budget locks: write the name and the result beside the line, then move the bill and the decision together, or move neither.

LeadershipOS™ walks through Decision Boundaries in full: who owns which decisions, under what conditions, and when to escalate, which is the half of a chargeback nobody writes down: https://TheLeadershipOSBook.com

Why "you're empowered to decide" makes a capable team freeze, and the one definition that lets them act without asking.

Anthony S. Jackson

Anthony S. Jackson

Anthony S. Jackson has spent 30 years inside technical organizations. He is the author of the Architecture Protocol Series: three books on the structural problems technical leaders were never told they would face. He writes the LeadershipOS™ Inner Circle, a monthly printed newsletter for CTOs and engineering managers who design teams that hold under pressure.

LinkedIn logo icon
Back to Blog