A number the board can read: reporting AI value without dressing it up

The board is not asking for a tour of the stack. It is asking whether the AI already on the books returned anything operating partners and portfolio-company CFOs can stand behind. Usage counts, a vendor’s unlabeled saving, and a slide of model names will not answer that question. A short pack will, if it is built from the same numbers finance uses everywhere else.

This note is about that pack: what belongs on a one-page board or LP summary, what should stay off it, and how to keep a projected figure from being read as a measured one. It is not marketing measurement, and it is not a method for crediting campaigns or channels. It is a reporting problem. The aim is a number the board can read, with the limits still attached.

Why the pack is now a funding question

Boards are asking for a return, and many finance teams do not yet have one they can defend.

A 2026 CloudZero survey of 260 senior finance leaders, 135 of them CFOs, reports that only 22% can tie AI spend to business outcomes. Eighty-seven percent say they need to do that within the year. Two-thirds of boards (66%) now condition further AI funding on proof of return. Twenty-two percent will approve nothing new until a return is shown, and only 26% still say, in effect, invest now and sort out the returns later. Forty-three percent of finance leaders in the survey are already being asked for a number they cannot give. CloudZero is a software vendor, and the figures are from its own survey. The report also states that the percentages describe associations in a single point-in-time survey, not proven cause and effect.

A separate 2026 Coupa survey of 600 global finance leaders, covering North America, Europe, Japan, and Australia, reports that 76% say difficulty quantifying AI ROI is hindering further investment. Coupa’s FAQ on the same page says that difficulty is moderately to significantly hindering further AI investment. Coupa is also a software vendor. The public landing page does not state fieldwork dates or a full method.

Taken together, the surveys describe a reporting gap, not a verdict on any one company, and they are not a reason to invent a return so that funding can continue.

What does not belong on the page

A one-pager fails when it asks the board to treat activity as a result.

Seats provisioned, prompts run, tokens consumed, and tickets touched are useful operating facts. They are not a return. They belong in a footnote or an appendix if they help explain volume, not in the line that answers what the spend produced.

Unlabeled vendor savings belong off the page as well. A projected reduction from a business case, a benchmark from a sales deck, or a “typical” productivity gain is not a result until finance has compared a stated “before” with what it can observe, and has said how sure it is. If the only number you have is the vendor’s, say that, and keep it out of the result column.

A pack with no stops and no labels will be read as a narrative, not as a count. If every initiative is presented as a success, or if every figure is presented with the same confidence, the board has no way to tell a measured cost-out from a hope. Include the initiatives you would pause or not renew. The absence of a stop is itself information, and it is usually the wrong information.

Do not add up unlike claims. Hours freed, quality improvements, and a possible avoided hire are different kinds of result. If you cannot show that the same hour produced more than one of them, do not count it twice. If you cannot yet translate an operational change into a P&L line, do not put a dollar figure in the headline and explain the gap in the fine print.

What belongs on the page

A short pack is enough. One page is the right length for a board or LP summary, because it forces a choice about what is actually known.

For the period in view, the page should show:

The inventory of AI spend in scope, by initiative, all in: subscriptions, usage, implementation, and the internal time you can reasonably assign to standing the work up and running it.
For each initiative, the result against a stated baseline — or a clear statement that a baseline was never recorded and has not yet been reconstructed.
A confidence label on each result: measured, estimated, projected, or not yet attributable.
A keep, fix, or stop recommendation, with one line of reason.
The limits of the data: the period, the sites or teams in the comparison, other changes in the business, and what you could not see.
What will be measured the same way next period, and who in finance owns that work.

Lead with outcomes, not model names. The first page should read like a cost-and-result schedule, because that is what it is.

“Not yet attributable” is a valid line, not a failure of the finance team. It is the right answer when rollout dates were not kept, when the only “before” is a recollection, when too many other changes hit the same workflow, or when all you have is a vendor’s projected saving. Put it on the page. It tells the board what not to put in an LP letter, and what to instrument before the next renewal.

If hours were freed and you cannot show a cost out, an avoided spend that would otherwise have been approved, or a volume increase that used the capacity, say that the capacity was released and that the financial translation is incomplete. Do not convert the hours to a salary-equivalent EBITDA figure and hope no one asks where the money went.

Write each line so finance can defend it

The one-pager is a cut of a ledger, not a different story. How TribuTrace builds the ledger is the longer method; the reporting rule is simpler. Every line the board sees should already exist, with its source and its label, in the working schedule finance owns.

For each initiative, write one sentence the finance owner can defend out loud:

1. What it cost in the period, and what is included in that cost. 2. What operational change you observed, and in which system. 3. What financial translation, if any, you can make from that change. 4. How you compared “before” and “after,” or why you could not. 5. The label, and the keep, fix, or stop call.

Keep the supporting detail close enough that a skeptical reader can reconstruct the claim: the window, the unit of work, the comparison group or the reason there isn’t one, and the other changes you ruled in or out. If you use an estimate, say which inputs are observed and which are assumed. If you use a projection, keep it in the projection column. Do not strip a label to make a number look better. A smaller measured figure is more useful than a larger unlabeled one.

The same page should also say what is out of scope. A baseline on two accounts-payable tools is not a company-wide AI return. A result at two sites is not a result at four. Naming the boundary is part of making the number readable.

A hypothetical one-page pack

The following is a worked illustration only. It is not a client result, and the figures are invented to show the shape of the page.

Suppose a portfolio company has three AI lines already in use: an invoice-coding tool in accounts payable, a general-purpose copilot, and a customer-service drafting assistant. Finance can see the subscriptions and the implementation invoices. It has not yet taken a number to the board.

The one-pager might read, in substance, as follows.

On accounts payable, cost for the last two quarters is stated in full. Cycle time and exception rates at the two early sites moved after go-live in a way that did not appear at the two later sites in the same window. Volume was stable, and no other AP redesign ran in parallel. The cost-per-invoice change at the early sites is labeled measured, with the limits stated: two sites, two quarters, this workflow only. Hours no longer spent on coding have not shown up as a headcount reduction; some went to exception review. Any earnings claim on the remaining hours stays off the page, or is labeled estimated and marked incomplete. The call is to keep the tool at the early sites, fix the exception load before expanding, and not claim a company-wide return.

On the copilot, cost is clear and the operational result is not. Seat counts and weekly active use are available. There is no agreed unit of work, no baseline, and no comparison group. The line is not yet attributable. The call is to instrument a small number of workflows before renewal, or to stop buying seats that have no owner.

On the service assistant, handle time is down in the months after rollout, but a staffing change and a new ticketing system landed in the same window. A dollar translation would be a guess. The operational movement can be reported as estimated, with the confounding changes named, and no EBITDA figure attached. The call is to fix the comparison, not to dress the estimate up as a measured saving.

None of those numbers should be copied out of this illustration. Only the shape of the pack should be: cost, result, label, decision, and limits, on one page, with unlike claims left unlike.

Who has to stand behind it

The operating partner needs a pack that can be compared, carefully, across companies, without forcing a roll-up of figures that were not measured the same way. The portfolio CFO needs a pack that will survive a question from the auditor of the numbers, even though this work is not an audit opinion. Those are different seats and the same standard: a figure without a label should not leave the room.

If the honest one-pager is mostly “not yet attributable,” that is still a board-ready result. It is a decision input. It tells the room what not to fund on the strength of a story, and what has to be true before the next purchase or the next renewal.

If you want a structured pass at this for one portfolio company, the AI Value Baseline is the fixed-scope version of the same work: a handful of initiatives already in use, a reconstructed baseline, a labeled ledger, and a short readout your team can take into the next review.

Sources

Figures were checked against the live pages on November 10, 2026.

CloudZero, “Finding the ROI of AI: The Finance Perspective,” 2026 survey of 260 senior finance leaders, 135 of them CFOs. The report states that 22% can tie AI spend to outcomes, 87% say they must within the year, 66% of boards condition further AI funding on proof of return, 22% will approve nothing new until a return is shown, 26% still favor investing now and sorting out returns later, and 43% of finance leaders are already being asked for a number they cannot give. The report also states that figures describe associations in a single point-in-time survey, not proven cause and effect. https://www.cloudzero.com/finance-needs-ai-roi-2026-survey-report/
Coupa, “The Strategic CFO Report 2026,” a survey of 600 global finance leaders in North America, Europe, Japan, and Australia. The public page reports that 76% say difficulty quantifying AI ROI is hindering further investment. The FAQ on the same page says the difficulty is moderately to significantly hindering further AI investment. Fieldwork dates were not shown on the page. https://www.coupa.com/resources/the-strategic-cfo-report/