Measuring Microsoft 365 Copilot ROI: what the Copilot Dashboard’s assisted hours do and don’t tell you

Roiva wrote this, and Roiva connects to Microsoft 365 Copilot. The facts about Microsoft's tools and the UK government trials were gathered from Microsoft's and gov.uk's published pages on October 24, 2026, and each one links to the page it came from. Microsoft revises these pages often, so check the link before you rely on a figure.

How do you measure the ROI of Microsoft 365 Copilot?

Set what the licenses cost against the work they gave back, measured on your own tasks rather than estimated from usage. Microsoft's Copilot Dashboard tells you who uses Copilot and estimates the hours it assisted. It doesn't measure your team's work, and it doesn't tell you whether any hour became cash.

That leaves three numbers to put in front of a CFO, each from a different place:

What is Copilot assisted hours?

Copilot assisted hours is Microsoft's estimate of the time Copilot helped employees over the last 28 days. It is worked out from counts of Copilot actions, each multiplied by a factor from Microsoft's research, not timed on your work. Microsoft calls the results "broad approximations based on the best available research rather than precise calculations" (Copilot Dashboard docs).

The method, as Microsoft documents it:

Microsoft adds that creation "should be in particular understood to be a broad estimate" (Copilot Dashboard docs).

What is Copilot assisted value?

Copilot assisted value is assisted hours multiplied by an hourly rate. The default rate is $72 an hour, from U.S. Bureau of Labor Statistics data, and a viewer can change the rate and currency in the dashboard's settings (Copilot Dashboard docs).

So the dollar figure on the dashboard is an action count, times a minutes factor from someone else's study, times a national average wage. Each of the three can be defended as a starting point. None of them is a measurement of your firm.

Why is assisted hours a weak number to give a CFO?

Because a CFO asks what changed in the business, and assisted hours counts what people clicked. Four things stand between the two:

• The minutes aren't yours. Six minutes per action comes from Microsoft's studies of knowledge workers, not from timing your people on your tasks.
• An hour assisted isn't an hour freed. A drafted email still gets read and edited. Microsoft's method credits the action, not the time the person actually got back.
• An hour freed isn't cash. Freed time becomes money only if it is billed, redeployed to work that earns, or used to avoid a hire. The dashboard can't see which.
• The history had a known error. Microsoft documents an issue that understated assisted hours, assisted value and two email metrics from June 1, 2025, through February 18, 2026. It says the fix applies from February 18, 2026, and that the history "will be restored as soon as possible" (Copilot Dashboard docs, as of October 24, 2026). A year-over-year comparison spanning those months needs that caveat.

None of this makes the dashboard wrong for its purpose. It is built to show adoption and where to push it, and it does that well.

What licenses does the Copilot Dashboard need?

None beyond Microsoft 365. Microsoft says that "neither a paid Viva Insights license nor a Microsoft Copilot license is required to view the dashboard," for any business or enterprise Microsoft 365 or Office 365 customer with Exchange Online (Copilot Dashboard docs, as of October 24, 2026). It analyzes the employees who hold a Copilot license.

What you see depends on how many licenses the tenant has:

The dashboard shows the previous 28 days, refreshed daily with a delay of up to six days. Microsoft's environment requirements page still says 50 Viva Insights licenses are needed for the dashboard and advanced analysis (environment requirements). The dashboard's own page is more recent and more specific, so it is the one quoted above.

What does the Copilot business impact report need?

The Copilot business impact report is a Viva Insights Power BI template that compares the business outcomes of heavier and lighter Copilot users. You upload the outcomes yourself, one row per person per month, up to five metrics, and the people measured need both a Viva Insights and a Copilot license. Microsoft says organizations "should include at least 65 Copilot and 65 non-Copilot users" (business impact report docs, as of October 24, 2026).

Microsoft is candid about what it shows: the differences "have not been evaluated for statistical significance," and Copilot use "is unlikely to be the only variable" behind them, with role and seasonality named as others. Heavier users may simply be the people who were already doing well.

It also has a practical catch. A firm that licensed every professional has no non-Copilot group of the same people to compare against, and comparing licensed professionals with unlicensed support staff compares two different jobs.

What did the UK government's Copilot trials find?

Two UK government studies are the largest public trials of Microsoft 365 Copilot, and they point the same way: people report saving time, and nobody has yet shown the time turned into more output.

Microsoft also commissions Forrester's Total Economic Impact studies of Copilot. The March 2025 enterprise study models a 116% ROI for a composite 25,000-person organization, and each study says it "is commissioned by Microsoft and delivered by Forrester Consulting" (Forrester TEI). A modeled composite is useful for building a business case. It isn't evidence of your return.

What does Microsoft 365 Copilot cost?

Microsoft 365 Copilot lists at $30 per user a month paid yearly, and Copilot Business at $18 per user a month paid yearly under a promotion that ends on December 31, 2026, as of October 24, 2026:

At $30 a month, 300 licenses cost $108,000 a year. That is the cost line cash ROI is taken against. The timed hours Copilot frees are a separate figure, reported beside it: they count against the $108,000 only where a document shows they became cash.

How should you measure Copilot for a CFO?

Measure a few named tasks properly instead of every action roughly. A method a finance team will accept:

1. Take the cost from the invoice. Copilot's part of the Microsoft bill, not the whole bill, if it carries other licenses too.
2. Count licenses against use each month. Licensed users beside the people who actually used Copilot in the month. Seats nobody uses are a cost to cut at renewal. Cutting them lowers the cost line; don't also count it as value.
3. Time a few tasks yourself. Pick three to five tasks people do often, such as first drafts, meeting notes or summarizing long documents. Time the same people doing them with Copilot and without, on days chosen at random so nobody picks their good days for one side.
4. Value only what you timed. Users × tasks per user a month × minutes saved per task ÷ 60 × a loaded hourly rate you can defend. For example, 150 users × 20 tasks × 5 minutes ÷ 60 × $50 an hour is $12,500 of freed capacity for the month. Ignore everything else people use Copilot for. It understates the value, which is the side a CFO would rather err on.
5. Report freed hours beside cash, never in it. Count freed hours as cash only where a document shows they became cash: billed work, a role not backfilled, overtime that stopped. ROI is taken over cash alone.
6. Write the rules down before the results. Which tasks, which rate, which months, signed by finance, so nobody moves the method after seeing the answer.

Keep the dashboard. It is the right tool for adoption, for finding teams that haven't started, and for deciding where training goes.

How does Roiva measure Microsoft 365 Copilot?

Roiva runs that method as a monthly routine, with the counting done for you.

• Use, against licenses. Connect Microsoft 365 Copilot and Roiva reads Microsoft's Copilot usage report every night: who holds a license and the last day each person used Copilot. It replaces each name with a digest it makes before anything is saved, and counts each month's licensed and active users. History starts with the first sync, because Microsoft's report gives only each person's latest day.
• Your own minutes. A time study in Roiva times one task with the AI and without it, assigns each participant's days without it at random, and publishes the means as the formula's baseline and reading. Roiva supplies no minutes of its own, so nothing is valued until your study sets them.
• Value as freed capacity. The AI Assistant Rollout template multiplies the month's active users by the timed tasks each does and the minutes saved on each, turns the minutes into hours, and prices them at your blended hourly rate. The result is reported as freed capacity, beside cash value and never added to it, and counts only once it is approved, by a person or by an auto-approve setting the account turns on.
• The rules, signed first. A measurement charter records the counting rules the CFO signs before any results exist.
• Cost from your books. Copilot's license cost comes from Microsoft's bill in QuickBooks, Xero or NetSuite, or a recurring cost you set on the initiative. Roiva doesn't book cost from the usage report.

Roiva doesn't read Microsoft's assisted hours or assisted value, and it doesn't see what anyone asked Copilot or what it wrote. Measuring a workplace AI assistant covers the same method for Claude Enterprise, ChatGPT Enterprise, Gemini in Google Workspace and Glean, and Amazon Quick's seats, and the AI ROI tools compared note covers the other tools a team weighing this usually looks at.