Shared AI Bills: Show the Rule Behind Each Team’s Cost

Two teams use the same AI workspace. One uses it to draft a newsletter; the other uses it to prepare customer support replies. At the end of the month, the provider sends one bill. The owner asks which team cost more to run. A dashboard can produce a tidy answer, but part of that answer depends on a rule the business chose.

That rule matters when a team uses the report to defend its budget, compare workflows or decide what to stop. You can divide a shared bill accurately and still give readers the wrong impression about what each team used. Start by separating the amount charged from the decision about who carries it.

Put the direct charges and the shared fee on separate lines

Consider a fictional studio. Its invoice shows a £120 workspace fee plus £40 in usage charges identifiable to the newsletter and £80 identifiable to customer support. The invoice totals £240. These figures are illustrative; they do not represent a provider’s prices.

The direct charges already have an owner in this example. The shared fee is the open question. If the owner wants each team to contribute equally to access, allocate £60 to each. The newsletter’s reported cost becomes £100 and support’s becomes £140. If leadership wants to fund access centrally, report £40 and £80 against the teams and leave £120 on a separate central line. Both views reconcile to £240.

The FinOps Foundation’s allocation guidance treats central funding as a deliberate option and describes splitting shared costs by fixed amounts, proportions or proxy measures. Its distinction is useful for a small studio: choose which costs teams should see and explain why, rather than treating full distribution as the only acceptable result.

Try two reasonable splits before trusting the ranking

Suppose the studio uses request counts as a stand-in for use of the shared workspace. The newsletter submitted 80 of 100 requests, so it receives £96 of the fee. Support receives £24. Including direct charges, the newsletter now shows £136 and support £104. Under that rule, the newsletter appears to cost more.

Now suppose the studio instead uses recorded minutes of active workspace use, with a quarter attributed to the newsletter and three quarters to support. The split becomes £30 and £90, and the reported totals become £70 and £170. Support appears to cost more. The invoice and the direct charges have stayed exactly the same.

Neither split measures what caused the provider to charge £120. Each answers a different internal question. Request counts give each request the same weight. Active minutes give greater weight to longer sessions. An equal split represents equal responsibility for access. The right starting point is the decision the report is meant to support.

The FinOps Foundation’s shared cloud cost guide distinguishes even splits from proportions held fixed or recalculated over time. For the fictional studio, a reasonable rule could be simple enough to explain in a sentence and stable enough to compare across months. More detailed tracking is useful only if it helps someone make a better decision.

Keep reporting and savings estimates apart

Suppose leadership sees £136 next to the newsletter and considers pausing it. If the £120 workspace fee remains payable and support continues using it, the £96 allocated to the newsletter does not disappear when the newsletter stops. Even the £40 direct charge is only a starting point for estimating savings: check whether the corresponding usage really stops and whether the invoice changes under the actual agreement.

For a budget review, show both questions explicitly: how much cost did the policy assign to the workflow, and which charges would change under the proposed action? The first can help explain responsibility; the second requires checking the billing terms and what would still run. MaryChuks.com’s guide to assessing AI cache savings makes a related point about keeping a change in one part of the bill in view alongside the whole workload.

Give unknown usage somewhere to go

Imagine next month’s usage export has 100 requests, of which ten have no team label. If the owner splits the fee using only the other 90, the percentages describe labelled requests. They might give a distorted picture if those ten requests came mainly from one team. An honest report can show the share left unresolved, or keep that share centrally funded for the period under a documented rule.

Do not count the same fee twice by showing the entire central amount in a total that already includes the redistributed amounts. Keep a check that direct charges, distributed shares and the remaining central or unassigned balance add back to the invoice on the same basis. A mismatch should be investigated before the figures are used to compare teams.

Write the rule next to the number

A short note beside the monthly report is enough to start. Make sure a colleague can answer these questions without opening a long spreadsheet formula:

• Which charges are direct? Name the record or label that supports the assignment.
• Which charges are shared? Identify the fee, period and amount being divided.
• What drives the split? Record the measure, its limitations and how missing records are handled.
• When can the rule change? Name an owner and show the change when comparing months.
• What does the report help decide? Distinguish budget responsibility from the charges that could actually be avoided.

If two plausible rules reverse the ranking, show both before making a consequential decision. That sensitivity is useful information about the report. The invoice tells you what the studio paid; the allocation tells you how it chose to share responsibility. Readers need both.

Sources

• FinOps Foundation: Allocation.
• FinOps Foundation: Managing Shared Cloud Costs.

Live guidance checked on 5 November 2026. All studio figures and scenarios are fictional.