Finance already knows the question it needs answered about Copilot: what came back for the money we spent? That isn’t the hard part. The hard part is producing a number that answers it on finance’s own terms, and on most Copilot ROI slides the word “return” is doing work it can’t carry.
To a finance person, a return is specific: revenue that went up or cost that went down, against money spent. Put a Copilot “return” of 75x in front of them (the headline multiple in our worked example) and you’ve made a claim in their language, so they test it in their language. The test is simple and it’s fatal: show me the line that moved. If the only thing behind the number is hours people might spend on something else, it isn’t a return. It’s a productivity index wearing a return’s clothes.
This is the second correction in the series. It’s mostly about a word, but the word decides whether the number reads as measurement or marketing.
The symptom: a number that invites the one question you can’t answer
The hours-saved multiple is useful: it tells you Copilot is creating capacity, and roughly how much. Label it “ROI” or “return” and you’ve invited one question: what did it return? For freed hours the honest answer is “nothing yet, it created capacity.” Given after you called it a return, that answer sounds like backpedalling. Given because you labelled it a productivity index from the start, it sounds like rigour. Same facts, opposite credibility.
The number didn’t fail the review. The label set a test the number was never going to pass.
What we assumed, and what Microsoft itself says
Our first readout called it ROI, because that’s what everyone calls it. Then we found Microsoft making the argument against us. Microsoft’s guidance on measuring the value of Copilot Studio agents, under a heading called “The time-savings trap,” says that “claiming value based on theoretical time savings alone undermines credibility.” That’s the vendor itself warning that leading with theoretical hours will cost you the room. The honest move is to put that warning on your own page before anyone else raises it.
So we did. The readout quotes that line next to the figure it applies to, which concedes the objection first and signals that the other numbers meet the same standard.
The fix: a productivity index, with cash shown beside it
The hours multiple is presented as a productivity index: a measure of capacity created, explicitly not a cash return. Beside it sits the cash-only ROI from the previous post, so a reader sees the capacity and the money in one view and is never left to assume the first is the second.
We got this wrong inside our own tool first. The portal’s front page showed the index labelled “Return,” coloured green as if any value were good news. We caught it in a verification pass, relabelled it “Productivity index,” and made the colour conditional so a weak month doesn’t read as a win.
That’s the part IT owns. The business sets the standard (a return means a budget line moved); IT and CB5 make sure the reporting can’t quietly drift from it, down to the label and the colour on the front page.
When a freed hour actually becomes money
The useful question isn’t whether to count the hours. It’s what has to be true before an hour of capacity turns into cash. Five things:
- A budget line actually falls. A role you don’t backfill, a contractor you release, a licence you consolidate. Without this, nothing converted.
- The time is redeployed to something of value, not quietly reabsorbed into a slower pace or a longer inbox.
- The task was going to be done anyway. Time “saved” on work that only exists because the tool existed isn’t a saving.
- The output is used. A drafted document nobody sends saved no one anything.
- The saving is net of verification and rework, the time spent checking and fixing the AI’s output, which is never zero.
Everything the program measures tells you where to look for a conversion event. Only a budget line falling proves one happened, and that’s a fact about the business, not the telemetry.
Why the humble label wins
Calling it an index feels like giving something up. It’s the opposite: an honestly labelled index is a number you can stand behind in any room. A “return” that can’t name a line that moved holds up only until the second question.
The five tests are also a roadmap. Each is a place where capacity could become cash if a department leader decided to act: consolidate these licences, redeploy that freed time to a backlog that’s actually revenue. The index says how much capacity exists; the tests say where to turn it into money.
The general rule: a number is a return only when a budget line falls. Until then it’s a productivity index. Name it that, show the cash figure next to it, and quote the vendor’s own warning before a skeptic does.
Next week: even the index isn’t uniform. Some of the hours inside it rest on strong published evidence and some on almost none, and the honest version grades them.
CB5 Solutions helps organizations understand the layers of AI, get the most from what they buy, and build an AI program that pays back inside their IT standards. We start every engagement with one business problem and a governance baseline. Let’s pick yours.