Why does metered AI need governance?
Metered AI needs governance because usage-based pricing without controls is an open-ended commitment. A worker that consumes tokens as it works can, in principle, consume more than intended, and a finance leader is right to ask what bounds it.
The answer on ftes.ai is a set of controls built into the pricing model rather than layered on as policy: the wallet, the budgets, the hard stop, and the visibility. Each is described below with who controls it.
What is the organization wallet?
The organization wallet is the block of tokens a plan buys, shared by every Digital FTE the organization hires. There is one wallet, not one per role, so the total exposure is the plan, and hiring another FTE does not open another line.
The wallet is topped up by the plan cycle or by an explicit top-up, both of which are permissioned actions. Nothing else adds to it, and nothing draws from it except the metered work of the FTEs.
How do budgets work inside the wallet?
Inside the wallet, the manager sets a budget per Digital FTE in its Setup tab and a budget per mission when a bounded push is launched. The FTE plans its work within the budget and reports spend against it; when a budget is close, the morning brief says so before it is reached.
A Performance Ads Digital FTE, for example, has a spend budget for the ad platform separate from its token budget for its own work, and both are visible. Budgets are bounds, not targets; an FTE that finishes its work under budget stops.
- Wallet: the plan's block, one per organization
- Per-FTE budget: set in Setup, spend visible in the brief
- Per-mission budget: set at launch, progress and spend visible
- Platform spend: separate budgets for ads or other external spend
What is the hard stop?
The hard stop is the rule that an empty wallet pauses every Digital FTE. Nothing runs on credit, no work is done that the organization has not paid for, and the workspace shows the pause plainly. Work resumes where it stopped after a top-up or the next cycle.
This is the control that makes the model safe for a finance leader. The maximum exposure in any period is the wallet, and the wallet is set by a permissioned purchase.
Who can change what?
Plan changes and top-ups are permissioned to the organization's owner or admin roles, and are recorded in the audit trail with who made them and when. Per-FTE budgets are set by the FTE's manager within the wallet. Mission budgets are set by whoever launches the mission, again within the FTE's budget.
The audit trail means spend governance is reviewable after the fact: who set the budget, who changed it, what was consumed. That is the same standard a finance function applies to any other budget line, applied to the AI line.
What does spend governance look like in a finance leader's month?
In a finance leader's month, spend governance is three short reads and one decision. Each week they glance at the morning brief's spend line for each role, which shows tokens against the per-FTE budget and, for roles that control external spend such as ads, that spend against its own cap, with warnings as either approaches its limit. At the monthly review they read usage per role and cost per outcome from the Targets record, and the FTE's own report of what it delivered for what it consumed.
The decision is the block for the next cycle, sized from the record rather than a forecast, and, where a role's cost per outcome is rising with flat outcomes, a question about its model tier or its remit. Plan changes and top-ups are permissioned to owner and admin roles and logged, so the audit trail shows who sized the block and when. The hard stop means the maximum exposure in any period is the block already purchased, which is why the month contains no surprise: the ceiling was set by a decision the finance leader made, and the record shows how close the work came to it.