AI adoption strategy on Microsoft starts with scope
Imagine the board asks, "Are people really using Copilot, and is it worth it?" You open the admin center, see a healthy count of users, and realize that number answers only half of the question. Any AI adoption strategy on Microsoft tools reaches this moment.
That's the trap in an AI adoption strategy on Microsoft: license and activity reports feel like proof. They're a start, not the answer. We'd treat them as the first line of the evidence, never the whole page.
Microsoft's documentation lists four reporting sources for Copilot: the admin center, Viva Insights Copilot Analytics, Purview audit logs, and Power Platform or Copilot Studio. The admin center reports focus on license eligibility, adoption and basic usage, per a Microsoft Learn page last updated 2025-09-17. Its usage view covers users who accessed Copilot, trends, apps used and top agents. That's activity data.
Done when you can list each report you rely on, its scope and who it counts.
Sources: Microsoft Copilot reports for IT admins | Microsoft Learn
Document who the numbers leave out
Before you read an adoption rate, ask: "Who isn't in this number?" Microsoft's documentation says the Copilot Dashboard analyzes only employees with an assigned Copilot license, so unlicensed staff don't appear in its adoption and impact view. Copilot Chat users without a license are covered separately in readiness and adoption insights.
Timing matters too. Per Microsoft's documentation, readiness, adoption and impact data cover the previous 28 days, with a delay of up to six days. Sentiment data follows the survey period you select.
Then check for known issues. Microsoft documented a problem that underestimated some email-related metrics, including Copilot assisted hours and Copilot assisted value, from June 1, 2025 through February 18, 2026. If you built a business case on those figures, recheck it. Microsoft says historical data will be restored, so confirm current status in Service health.
In our view, a clean-looking dashboard deserves more suspicion, not less, because gaps in scope never show up as errors.
Sources: Connect to the Microsoft Copilot Dashboard for Microsoft 365 customers | Microsoft Learn
Label every metric as activity, perception or outcome
Copilot's sentiment metrics draw on Pulse and Glint survey data, according to Microsoft's documentation, and are listed for tenants with 50 or more Copilot or Viva Insights licenses. Microsoft also describes the Impact tab as combining qualitative and quantitative survey data on how users feel about Copilot's impact. Those are perceptions, collected in Microsoft's own surveys.
Outcomes are different. Advanced Reporting can combine Copilot metrics with business outcome data such as ticket resolution time or case resolution rate, but you have to upload that data yourself. The tool shows the two side by side; it doesn't prove one caused the other.
Analyst Kevin Kieller, writing in No Jitter in March 2026, argues that productivity or quality gains attributable to Copilot are hard to see and measure, which makes ROI hard to calculate.
Our suggestion: build a one-page metric ledger. Tag each number as licensed, active, perceived or business outcome. You're done when no metric is unlabeled.
| Label | Example | Source of the number |
|---|---|---|
| Licensed | Seats assigned | Admin center |
| Active | Users who accessed Copilot | Usage reports |
| Perceived | Reported sentiment about impact | Pulse or Glint, or your own survey |
| Business outcome | Case resolution rate | Your uploaded data |
Sources: Connect to the Microsoft Copilot Dashboard for Microsoft 365 customers | Microsoft Learn, Copilot controls measurement and reporting | Microsoft Learn, Copilot Analytics introduction | Microsoft Learn, 4 obstacles impede paid Microsoft 365 Copilot adoption
Ask what usage logs can't answer
Logs show that someone opened Copilot. They don't show whether that person feels capable, what blocked their colleague, or who thinks it's worth the effort.
In an exploratory interview study of 10 experienced U.S. professionals using Microsoft 365 Copilot (September to December 2024), none named formal training as their main way of learning. Eight relied on trial and error and six on tips from colleagues.
The same group reported perceived efficiency gains but low confidence with advanced features. It's a tiny, non-representative sample, so read it as a prompt for questions, not a finding about your people.
In Microsoft's Work Trend Index 2026 survey of 20,000 AI users in 10 markets, fielded February 18 to April 20, 2026, only 26% said their leadership is clearly and consistently aligned on AI. The survey associated culture and manager support more strongly with self-reported AI impact than individual factors. These are self-reported associations, not proof of cause, and describe those AI users, not your company.
Use them to decide what to ask, not what to conclude. Suggested survey topics (our suggestions, not a validated instrument):
- How confident are you using Copilot for your core tasks?
- What gets in the way: access, rules, time, know-how or support?
- Does your manager use AI in front of the team?
- Are people rewarded for trying new ways of working?
- How useful is Copilot to your actual work?
Sources: 2026 Work Trend Index report: Agents, human agency, and opportunity, Beyond Training: How Workers Discover Value in Enterprise AI
Tie adoption to outcomes without over-claiming
Pick two or three business outcome metrics per function before rollout. A support team might track case resolution; a service desk, ticket time. If you use Advanced Reporting, upload them, and treat any pattern as something to investigate.
A separate Microsoft-led study of 1,800 workers worldwide, reported second-hand in the Work Trend Index 2026 page, linked managers modeling AI use to a 17-point lift in reported AI value. That justifies asking about manager behavior.
We'd skip any claim that licenses plus activity equal value. Report perceived value as perception, and let outcome data speak only for itself.
Which items above are Microsoft-documented facts and which are our suggestions? The reports, scopes, windows and known issues come from Microsoft's documentation. The ledger, survey topics and outcome-picking steps are ours.
Sources: Copilot Analytics introduction | Microsoft Learn, 2026 Work Trend Index report: Agents, human agency, and opportunity
Add a survey, then repeat it
Your Microsoft reports tell you who has a seat and what they touched. A short employee survey tells you whether people have the skills, the permission and the time. A complete AI adoption strategy on Microsoft needs both.
Run the survey on a regular cadence, so you can see change and choose between training, access, manager coaching or governance.
AI Adoption Score uses predefined employee surveys to measure usage, skills, barriers, culture and perceived impact, and to track change over time. It doesn't read logs, and it reports perceived impact. Start with the ledger this week, then add the survey that fills its empty rows.
Frequently asked questions
Who should own the metric ledger?
Give it to whoever owns the AI adoption strategy, usually a Head of AI, CIO or digital transformation lead. They should review it with IT, HR and function heads, since each supplies different numbers. It's our suggestion, not a Microsoft requirement.
How often should we survey employees about Copilot?
Quarterly is a sensible starting rhythm, though that's editorial guidance. It's frequent enough to see change after training or manager coaching, and infrequent enough to avoid survey fatigue. Keep the questions stable between rounds so results can be compared.
Can Copilot's own sentiment data replace a separate survey?
Not fully. Microsoft lists its survey-based sentiment metrics only for tenants with 50 or more Copilot or Viva Insights licenses, and they rest on its own surveys. For your specific skills and barrier questions, you'll likely need a survey of your own.
Should unlicensed employees be surveyed too?
Yes, if you want to understand barriers to access. The Copilot Dashboard analyzes only employees with an assigned license, so unlicensed staff are invisible there. Surveying them shows whether lack of access, rules or interest is holding them back.
Sources: Connect to the Microsoft Copilot Dashboard for Microsoft 365 customers | Microsoft Learn
