Everyone in the room agrees the AI bill is too big, and each team places the line item with another team.
Finance treats it as a consumption decision that engineering controls, and engineering treats it as a budgeting decision that finance owns. FinOps says their charter covers cloud infrastructure, and this is neither. The meeting ends without a decision, and the spend keeps compounding while the org chart gets debated.
One product leader we heard from described the end state this produces. They wanted an executive view that ties a decision like a workforce reallocation back to whether the outcomes and the cost are landing where they expected, and today that view doesn't exist. In their words, "right now everything's sort of opaque to everyone."
A cost stays opaque when no one has been assigned responsibility for clearing it up. An owner carries the discomfort that keeps a number under regular question, and this line currently has no owner.
The reasonable version of the objection
Each team is right that the decision isn't theirs alone to make, and the standoff persists even when everyone is acting in good faith.
AI spend sits in more than one domain at once. Engineering controls the consumption, finance owns the budget, and the variable, cloud-adjacent shape also makes it look like a FinOps cost. Each team can correctly say the decision requires the others. That's a real coordination problem, and telling people to just pick someone doesn't solve it.
AI breaks the normal pattern for who owns a system. Most infrastructure has one obvious owner, the person whose remit it already sits in, so a change to it is theirs to make. A traditional integration platform belongs to whoever runs integration, and that settles the question. AI spreads across every function at once, so agreeing to do anything with it pulls in everyone, and it usually ends up with a committee because no one function can claim it the same way.
Why AI spend lands in no-mans-land
Several forces push it there, and none of them resolve on their own.
It arrived as a substitute for headcount, so it gets managed like headcount. Most functional groups run on a fixed annual budget, and most of that budget is headcount. As teams adopt coding assistants and other AI tools, their inference spend climbs against a budget that's supposed to stay neutral, which forces tradeoffs. A cost that has to stay budget neutral against headcount ends up with whoever controls headcount, and that person usually sits far from the teams generating the usage.
The classification problem runs deep enough that workforce-planning software is now being extended to model agents as a new kind of resource, so their cost can be compared against traditional roles. When the tools built for budgeting have to change to hold AI, the ambiguity isn't confined to any one company.
Adoption moved faster than governance. Agents went into production under a broad mandate to move fast on AI, ahead of any process for governing them. Teams that moved quickly are now realizing they have no governance model for agents at all, and deferring that question left ownership unassigned while the spend went live.
Taking the line item carries real exposure, which is part of why it stays unclaimed. Whoever takes the line inherits the number and its growth rate, along with the eventual conversation about what it bought. That exposure is real when leadership has tied its own credibility to AI results, which makes the line hard to volunteer for. Pressure comes from the other direction too, with executives asked constantly how the company is using AI and expected to answer in outcomes rather than login counts.
That avoidance ends with a leadership team that has spent seven figures and can't say what value it got. The spend accumulates either way, and without an owner there's no one to account for it.
Who should own the line, in our view
Departments should own AI spend the same way they own headcount. A functional leader already answers for a fixed annual budget, and most of that budget is people. Inference spend now competes against those same dollars, so the person weighing another engineer against another coding assistant is the person who should carry the line.
That puts ownership where the spending decision gets made. A department head decides which tools get turned on and how heavily they're used, and the cost of those choices lands in their own budget rather than a central pool nobody feels. Finance keeps the role it plays for headcount, setting the envelope and consolidating the departmental lines into a total it can report. Engineering keeps the infrastructure and controls that make usage measurable. Neither one inherits accountability for spend it doesn't direct.
The finger-pointing from the top of the post ends because the tradeoff is contained. When AI spend and headcount draw from the same budget under the same owner, there's no other team to hand the decision to. Hiring one fewer role to fund an agent rollout becomes a call a department leader can make and defend.
What makes it work is attribution each department can trust. A blended invoice split by guesswork gives a leader nothing to manage against, so consumption has to resolve to the team that generated it, and those same numbers have to roll up for Finance to see the company total. Once every department watches its own spend against its own budget, the line has as many owners as there are budgets, which is how headcount already works.
Where we fit
A department leader taking on this line needs the same things in their first week, and so does the Finance team consolidating those lines.
They need to know who spent what, which is Attribute Spend to Teams & Departments. They need thresholds that warn before the quarter is decided, which is Set Budgets & Alerts. They need a ranked list of what to fix, which is what AI Insights produces, with each finding tied to potential monthly savings and a suggested action. They need a defensible view of return, which is Analyze ROI & Unit Economics. And they need the reporting to survive contact with a board deck, which is Build Custom Dashboards & Charts.
The org chart argument gets easier once somebody can show leadership what the spend bought.
