What Your Copilot Usage Data Means
7 adoption insights you can surface today within Copilot
Here’s something many organizations overlook: Microsoft 365 Copilot can actually analyze its own adoption data. By pulling usage patterns and engagement metrics from the Microsoft 365 admin portal, Copilot can show you which teams are getting the most value from AI, where adoption gaps exist, and what to do about it.
That kind of visibility gives leaders a real advantage.
Instead of guessing where to invest in training or enablement, you can make targeted decisions backed by actual usage data and start building a culture of AI-driven productivity across the organization.
How to access your Copilot usage data
Getting to your adoption data takes just a few steps. You’ll need AI Administrator access or higher in the Microsoft 365 admin portal.
- Navigate to the Microsoft 365 admin portal.
- In the left navigation pane, select Reports > Usage.
- Navigate to Microsoft 365 Copilot > Copilot.
- (Optional) Adjust the date range filter – options include 7, 30, 90, or 180 days.
- Scroll to the Copilot usage details section and click “Export.”
- Open the exported CSV file in Excel.
- Click the Copilot button to open the AI pane and select “Analyze this workbook and give me interesting data insights.”
- Review the output to identify patterns and opportunities for your organization.
What follows are real insights generated by Copilot from a sample dataset of 53 users. Think of this as a walkthrough of what your own analysis might look like and what to do with what you find.
1. Adoption is top-heavy
Across 53 users, the average number of prompts submitted was approximately 239, but the median was only 199, with a maximum of 871 and a minimum of zero. That “mean greater than median” pattern is a classic sign of a power-user distribution, where a small group of heavy users is pulling the average up.
Why it matters: This is where a champions program pays off fastest. The top cohort already knows what works as their best prompts and workflows are ready to be documented and scaled to the broader organization. Rather than building training from scratch, you’re capturing and replicating real-world success.
2. The top 10 users are far ahead of the pack
The top 10 users by total prompts submitted ranged from 406 to 871, which is well above the median level of 199. That’s a sizable group, large enough to form a meaningful cohort but small enough to interview quickly for repeatable patterns.
Why it matters: These users have all the hallmarks of early champions. They’re generating enough volume to have developed real habits and preferences, which makes them ideal founding members of a Copilot Center of Excellence. Start here when building peer-led enablement.
3. Work chat vs. web chat behavior splits sharply
At the dataset level, users averaged about 90 work chat prompts (median of 32) compared to roughly 55 web chat prompts (median of just 3).
That median of three in the web chat is telling: the majority of Copilot’s value comes from work-related prompts, which leverage the Microsoft Graph to pull from OneDrive, Teams, email, and other company-specific data. The top cohort shows the split even more clearly. Steve Martin submitted 722 work chat prompts and zero web chat. James Berman went the opposite direction with 48 work chat and 466 web chat. John Smith landed in between at 345 and 320, respectively.
Why it matters: It suggests two distinct adoption motions are at play: “Teams-native” users who lean on Copilot for day-to-day collaboration, and “research/ideation” users who rely on web chat for exploration and drafting. Recognizing this split allows you to tailor training for each group rather than taking a one-size-fits-all approach.
4. Prompts per active day reveals who’s really leaning in
Raw prompt counts don’t tell the whole story. When you normalize by active usage days, a clearer picture of intensity emerges.
Jane Doe averaged about 14.8 prompts per day over 59 active days. John Smith came in at 13.6 over 60 days, and Steve Martin at 12.7 over 58 days. Trevor Hartley is an interesting outlier: 11.8 prompts per day, but across only 18 days, making him a “burst” user who goes deep during specific projects.
Why it matters: Intensity metrics help you distinguish between habitual users and project-driven ones. Both are valuable but need different support. Steady users are great candidates for feedback loops and feature testing, while burst users may benefit from targeted enablement that helps them sustain engagement beyond individual projects.
5. Zero-usage and single-surface pockets exist
Across the dataset, two users had no prompts in any app, seven had no work chat prompts, and 25 had no web chat prompts. That last number is especially notable, as it shows that nearly half the user base has never used web chat.
Why it matters: When that many users show zero activity on a specific surface, it usually points to one of two things: they don’t see it as part of their workflow, or they don’t know when to use it. Either way, a short enablement nudge in the form of a few concrete examples and some “when to use web chat” guidance could unlock meaningful adoption gains with minimal effort.
6. Teams and work chat lead adoption while agent is still niche
Looking at which Copilot surfaces users have touched at least once, Teams and work chat lead the way — 51 out of 53 users have been active on each. The M365 Copilot app (49), Word (47), Outlook (47), and Excel (45) round out the most widely adopted surfaces. On the other end of the spectrum, OneNote Copilot has only been used by 9 users, Copilot Agent by 8, and Loop Copilot by 16.
Why it matters: Your adoption footprint is strongly Teams- and chat-centric, which makes sense given where most collaboration happens. The Agent numbers are worth watching but not alarming. They signal that Agent is still an early-adopter feature that is not yet ready for broad rollout. If increasing Agent usage is a priority, it’ll require targeted education and clear use-case demonstrations.
7. Recent activity shows where enablement will move the needle
Looking specifically at activity over the last 30 days, Teams Copilot leads with 46 recent users, followed closely by Copilot Chat (work) with 45. From there, the drop-off is significant: only 15 users have recently used Word Copilot, just 6 have used PowerPoint Copilot, and Copilot Agent has only 3 recent users.
Why it matters: This is essentially a roadmap for where enablement will have the biggest impact. Teams and chat are already humming, so incremental gains there will be harder to come by. PowerPoint and Word represent substantial untapped opportunities. Showing users end-to-end workflows like “meeting notes to document to presentation” could drive fresh engagement in the exact areas where content creation ROI lives.
Generating an executive summary
Once you’ve reviewed your adoption data, Copilot can take it a step further. Enter the prompt “Provide adoption reporting for leadership” and Copilot will generate a high-level summary you can share with stakeholders.
Here’s an example of what that summary looks like based on our sample dataset:
Turning insights into action
The overall picture from this data is clear: Copilot adoption is strong, but it’s uneven. A core group of power users is driving the majority of value, while a long tail of users shows light or inconsistent engagement. That’s not a failure. It’s an opportunity.
The most immediate move is to formalize a Copilot Champions Program. Your top 10–15 users already have the workflows, prompts, and habits worth replicating. Capture what they’re doing, package it into short internal playbooks, and let their success stories do the heavy lifting for peer-to-peer enablement.
From there, focus your training investments where the data says they’ll matter most. With 25 users at zero web chat activity, there’s a clear opportunity to show people when and how web chat adds value for research, analysis, and early-stage drafting. Similarly, PowerPoint and Word Copilot usage is well below its potential. Targeted campaigns that walk users through end-to-end content creation flows can unlock real productivity gains in areas people already work in daily.
Finally, keep Agent adoption in perspective. With only 8 users showing any activity and just 3 in the last 30 days, this is still early-adopter territory. Treat Agent as a focused pilot. Identify a few high-impact use cases, measure the results, and scale deliberately rather than pushing for broad rollout before the value case is proven.
Ready to turn your copilot data into strategy?
Inviso helps organizations go beyond surface-level adoption metrics to build real, measurable AI enablement strategies. Whether you’re standing up a champions program, optimizing Copilot workflows, or building the case for broader AI investment, we can help.
Get in touch at info@invisocorp.com to start the conversation.
