Every enterprise that deployed Microsoft 365 Copilot has the same experience six months later. The executive sponsor asks: “Is Copilot working?” The IT team says: “It’s deployed.” The CFO asks: “Is it worth $1.8 million a year?” Nobody has the data to answer. Deployment is not adoption. Adoption is not value. And value is not evenly distributed across 5,000 licensed users. In every Copilot deployment we have assessed, the pattern is the same: 15–25% of licensed users are power users who interact with Copilot daily and derive measurable productivity gains. Another 30–40% use it occasionally — a few prompts per week, mostly in Teams meetings and Outlook. And 35–50% of licensed users have used Copilot fewer than five times in the past month, or not at all. That bottom tier represents $126–$180 per user per year in wasted licensing. Across a 5,000-user deployment, the underutilised licences alone cost $630,000–$900,000 annually. This guide shows you how to find the usage data, interpret it, build the ROI case, and make the commercial decision: expand to more users, reassign from low-usage to high-potential users, or reduce the licence count at renewal.
Microsoft 365 Copilot is the most expensive per-user add-on Microsoft has ever launched. At $30/user/month, a 5,000-user deployment costs $1.8 million per year. A 20,000-user deployment costs $7.2 million. These numbers were approved based on projected productivity gains: time saved drafting documents, summarising meetings, analysing data, managing email. The projections were compelling. The reality is more complicated.
Copilot adoption follows a predictable curve. In the first two weeks, usage spikes as users experiment. Over the next 30 days, usage drops sharply as the novelty wears off and users return to established workflows. By day 90, the user base has stratified into three distinct groups that remain remarkably stable:
Power users (15–25% of licensed users): These users interact with Copilot 10+ times per day across multiple applications. They have integrated Copilot into their core workflows: drafting in Word, analysing in Excel, preparing presentations in PowerPoint, managing email in Outlook, and summarising meetings in Teams. For these users, Copilot delivers measurable time savings of 30–90 minutes per day. The $30/month licence is a clear investment.
Occasional users (30–40%): These users interact with Copilot a few times per week, primarily in one or two applications (typically Teams meeting summaries and Outlook email drafts). They find Copilot useful but have not fundamentally changed their workflows. The productivity gain is real but modest — perhaps 15–30 minutes per week. At $30/month, the ROI is marginal but defensible.
Non-users and minimal users (35–50%): These users have used Copilot fewer than five times in the past month, or not at all. Some tried it and found the outputs unhelpful for their specific work. Some never received adequate training. Some work in roles where Copilot’s current capabilities do not add value (data entry, process-driven tasks, roles that do not produce documents or attend meetings). For these users, the $30/month licence is a pure cost with no return.
The enterprise that does not measure usage cannot distinguish between these three groups. It pays $30/month for every user regardless of whether they use Copilot once a day or once a quarter. Usage tracking is not an administrative exercise. It is a $360-per-user-per-year commercial decision.
Microsoft provides several data sources for tracking Copilot adoption. None of them is perfect. Together, they provide a comprehensive picture.
The M365 admin centre includes built-in Copilot usage reports accessible to Global Administrators and Report Readers. These reports show adoption metrics at the tenant and user level: the number of users who have used Copilot in each application (Word, Excel, PowerPoint, Outlook, Teams, OneNote, Loop), the number of active users over 7-day, 30-day, and 180-day windows, and trend data showing whether adoption is growing or declining.
What it shows well: Aggregate adoption trends, application-level usage distribution, and the basic question of “how many users touched Copilot this month.” The admin centre reports are the fastest path to a high-level adoption assessment — they require no configuration and are available immediately.
What it does not show: The depth and quality of usage. A user who opened a Copilot prompt once and closed it without using the output counts the same as a user who generated 50 meaningful documents with Copilot in the same period. The admin centre reports show breadth (how many users), not depth (how much value).
Viva Insights (included in Microsoft 365 E5, or available as an add-on for E3) provides deeper Copilot analytics for organisations that have deployed it. The Copilot Dashboard in Viva Insights tracks usage patterns across the organisation, shows which applications drive the most Copilot interactions, identifies departments and teams with high and low adoption, and provides estimated time savings based on interaction patterns.
What it shows well: Departmental and team-level adoption patterns, which are essential for identifying where Copilot training is needed, where the tool is a natural fit, and where the licence investment is not generating returns. The estimated time savings metric (while imprecise) provides a directional indicator that helps build the business case.
What it requires: Viva Insights must be enabled and configured. Privacy controls must be set appropriately — Viva Insights can report at the individual user level, which raises employee privacy considerations that HR and legal should review. Many organisations configure Viva Insights to report at the team or department level (minimum group size of 10–25 users) to protect individual privacy while still providing actionable adoption data.
The Microsoft Graph API exposes Copilot usage data programmatically, allowing IT teams to build custom reports and dashboards. The Graph API provides more granular data than the admin centre reports, including per-user interaction counts, application-specific usage metrics, and the ability to correlate Copilot usage with other M365 activity data (meetings attended, emails sent, documents created).
What it enables: Custom Power BI dashboards that combine Copilot usage data with HR data (role, department, location, tenure) and business outcome data (sales revenue, tickets closed, documents produced). This correlation is where the ROI analysis moves from “people are using Copilot” to “Copilot usage is correlated with measurable business outcomes.”
What it requires: Technical effort to build and maintain the reporting pipeline. An analyst or developer familiar with the Graph API, Power BI, and data modelling. Most organisations underinvest in this capability and rely on the admin centre reports alone, which limits their ability to build a rigorous ROI case.
Quantitative usage data tells you who is using Copilot and how often. It does not tell you why non-users are not using it or what would make occasional users into power users. Structured user surveys at 30, 60, and 90 days after deployment provide the qualitative layer: which Copilot features are most valuable, which produce unhelpful outputs, what training gaps exist, and what workflow integration obstacles users encounter.
The critical insight from surveys: In most deployments, 40–60% of non-users cite “I don’t know what to use it for” or “the outputs weren’t good enough when I tried” as the primary reason for non-adoption. These are solvable problems — targeted training and use-case-specific prompt guidance can convert a significant portion of non-users into occasional or power users. A smaller percentage (10–20% of the licensed population) are in roles where Copilot simply does not add value today. These users are licence reallocation candidates.
Effective Copilot usage tracking requires a structured measurement framework, not ad hoc reporting. Here is the framework that produces actionable data:
Day 30 — Activation check: What percentage of licensed users have used Copilot at least once? If activation is below 70%, there is a deployment or communication problem — users may not know they have access, may not have received onboarding, or may face technical blockers. The day-30 check is a deployment quality metric, not an adoption metric.
Day 60 — Habit formation check: What percentage of activated users have used Copilot in 3 or more of the last 4 weeks? Weekly recurring usage indicates that Copilot is becoming part of the workflow. If fewer than 50% of activated users show weekly usage, the onboarding and training programme needs intervention. Deploy use-case-specific guidance, team-level champions, and department-specific prompt libraries.
Day 90 — Stratification assessment: By day 90, the user base has stabilised. Classify every licensed user into one of the three tiers: power user (10+ interactions/week), occasional user (1–9 interactions/week), or non-user/minimal user (<1 interaction/week over the past 30 days). This classification drives the commercial decision: reassign, train, or reduce.
Percentage of licensed users who have interacted with Copilot at least once in the past 30 days. Target: 70%+. Below 50% indicates a systemic adoption failure. This is the headline metric for every board-level Copilot report.
The percentage of users in each tier: power users (daily), occasional users (weekly), non-users (monthly or never). This distribution determines the effective per-unit ROI. An enterprise with 80% power users has a fundamentally different ROI profile from one with 20% power users and 50% non-users — even if the total licence count is identical.
Which applications drive the most Copilot usage? Teams meeting summaries and Outlook email drafts typically dominate early adoption. Word document generation and Excel data analysis indicate deeper workflow integration. PowerPoint presentation creation indicates advanced adoption. If 90% of usage is in Teams meeting summaries only, the enterprise is getting narrow value from a broad-capability tool — training should focus on expanding into Word, Excel, and PowerPoint use cases.
Which departments have the highest and lowest adoption rates? Marketing, sales, HR, and executive teams typically show the highest adoption (content creation, communication, and analysis are their core activities). Finance, engineering, and operations teams often show lower adoption (their work involves structured processes, specialised tools, and data formats that Copilot handles less effectively). Department-level variance informs both the training strategy and the licence allocation strategy.
Is usage growing, stable, or declining month-over-month? Growing adoption indicates the tool is finding product-market fit within the organisation. Stable adoption indicates the user base has stratified and further growth requires intervention (training, new use cases, management encouragement). Declining adoption is a red flag that demands immediate investigation — users are actively abandoning a tool they previously tried.
Total Copilot spend divided by the number of users who used Copilot at least weekly in the past 30 days. If 5,000 users are licensed at $30/month ($150,000/month total) and 2,500 users are weekly-active, the effective cost per active user is $60/month — double the licence price. This metric reframes the ROI question: is Copilot worth $60/month to the users who actually use it? The answer may be yes, but the enterprise should not pretend it is paying $30.
Usage data alone does not prove ROI. Usage proves adoption. ROI requires connecting adoption to business value. Here is the framework:
Microsoft’s published research claims Copilot saves users an average of 11 minutes per day. Enterprise-specific measurements typically show a wider range: 5–15 minutes/day for occasional users and 30–90 minutes/day for power users. The time savings model multiplies the measured time savings by the user’s hourly cost to produce a dollar value.
Example calculation: A power user saving 45 minutes/day at a fully loaded cost of $75/hour saves $56.25/day or approximately $1,238/month. The Copilot licence costs $30/month. The ROI for this user is 41:1. An occasional user saving 10 minutes/day saves $12.50/day or $275/month — still a positive 9:1 ROI. A non-user saving zero minutes provides zero return on the $30/month investment.
The aggregate calculation: For a 5,000-user deployment at $1.8M/year: if 1,000 power users save an average of $1,000/month, 1,750 occasional users save $250/month, and 2,250 non-users save nothing, the total monthly value is $1,437,500 against a cost of $150,000 — a 9.6:1 aggregate ROI. But the non-users represent $67,500/month ($810,000/year) in zero-return licensing. Reassigning those 2,250 licences to users in departments with proven high adoption would either reduce cost by $810K/year or redirect that value to users who would generate returns.
Time savings is the easiest ROI metric to calculate but not always the most important. Copilot also affects output quality: better-written proposals, more insightful data analysis, more polished presentations, faster customer response times. These quality improvements are harder to quantify but may represent more business value than time savings alone.
The practical approach: survey power users and ask them to estimate not just time saved but revenue influenced (deals where Copilot-assisted proposals played a role), errors avoided (data analysis mistakes caught by Copilot), and capacity created (projects that were only feasible because Copilot freed up time). Anecdotal evidence from 50–100 power users, structured and aggregated, produces a compelling qualitative case that complements the quantitative time savings data. For the complete ROI framework, see the Copilot ROI assessment and how to justify or challenge AI feature costs.
At 90+ days post-deployment, the usage data produces one of three commercial outcomes:
If adoption is strong (60%+ weekly active users), the ROI case is positive, and departments not yet licensed are requesting access, the correct decision is to expand the Copilot deployment. Negotiate the expansion within the EA framework — volume commitments for phased expansion can secure pricing protections for the additional licences. Document the ROI data from the initial deployment to strengthen the business case and the negotiation position. See negotiating Copilot pricing.
If adoption is moderate (40–60% weekly active users) with clear departmental variance, the correct decision is to maintain the total licence count but reassign licences from low-adoption users to high-potential users in departments with proven adoption patterns. This is the most common outcome and the most operationally complex.
The reassignment process: Identify users with fewer than 2 Copilot interactions in the past 60 days. Notify them that their Copilot licence will be reassigned in 30 days unless usage increases (this notification alone drives a 10–20% activation response from users who simply forgot the capability existed). After 30 days, reassign remaining unused licences to a waitlist of users in high-adoption departments. Repeat the measurement cycle.
The governance mechanism: Establish a quarterly Copilot licence review. Every 90 days, pull usage data, identify the bottom 10–20% of users by interaction frequency, and run the notification-and-reassignment cycle. This continuous optimisation prevents licence drift — the gradual accumulation of licences assigned to users who no longer use the product — which compounds over the EA term and wastes hundreds of thousands of dollars.
If adoption is low (below 40% weekly active users) despite adequate training and onboarding, the enterprise should reduce the Copilot licence count at the next available contractual opportunity. The reduction conversation with Microsoft is uncomfortable but commercially necessary. The data supports it: “We deployed Copilot to 5,000 users. After 6 months of active promotion and training, 2,000 users show regular adoption. We need to right-size to 2,500 licences (maintaining a growth buffer) and redirect the $900,000 annual savings to other priorities.”
Contractual considerations: Copilot licence commitments within an EA may have minimum terms. Check whether your EA allows mid-term reduction of Copilot licences or whether the reduction can only occur at renewal. If mid-term reduction is not contractually available, the reassignment strategy (Outcome 2) preserves the total licence count while maximising the value extracted from each licence. At renewal, the usage data provides the negotiation foundation for right-sizing. See negotiating Microsoft contract terms and negotiating termination and renewal options.
Before reducing licences, invest in converting non-users. The cheapest Copilot licence is one that is already purchased and starts generating value. The following interventions consistently improve adoption rates by 15–30 percentage points:
Generic Copilot training (“here is how to use Copilot in Word”) produces generic adoption. Role-specific prompt libraries (“here are 20 prompts for financial analysts in Excel,” “here are 15 prompts for HR managers in Outlook”) produce targeted adoption. Users who see exactly how Copilot applies to their daily work adopt at 2–3x the rate of users who receive only generic training.
Identify the power users in each department and formalise their role as Copilot champions. Give them 30 minutes in team meetings to demonstrate their most valuable Copilot workflows. Peer demonstration is more effective than IT-led training because it shows Copilot working in the context of real work, not abstract examples.
When department heads receive monthly reports showing their team’s Copilot adoption rate relative to the organisation average, adoption increases. Nobody wants to lead the department with the lowest AI adoption score. This is not about punishing non-users — it is about creating organisational awareness that Copilot is a strategic investment with expected returns, not an optional tool.
Teams meeting summarisation is consistently the Copilot feature with the highest adoption rate and satisfaction score. If a user is not using Copilot at all, the first intervention should be: “Turn on Copilot for your next three Teams meetings and read the summary.” Meeting summaries require zero behaviour change (the user just attends the meeting as normal) and deliver immediate, visible value. Once a user experiences the meeting summary, they are significantly more likely to explore Copilot in other applications.
Audit the technical deployment for blockers: are Copilot features enabled in all M365 applications? Are there conditional access policies or DLP rules that inadvertently block Copilot functionality? Are users on M365 app versions that fully support Copilot? In every deployment we have assessed, 5–10% of “non-users” are actually “blocked users” who cannot access Copilot due to configuration issues rather than choice.
Copilot licence management is not a one-time exercise. It is a recurring governance process that should be embedded in the organisation’s Software Asset Management (SAM) programme. The framework:
Monthly: Pull the M365 admin centre Copilot usage report. Flag any user with zero Copilot interactions in the past 30 days. Monitor the trend direction (growing, stable, declining).
Quarterly: Conduct the full stratification analysis: power users, occasional users, non-users. Run the reassignment cycle for users in the non-user tier for two consecutive quarters. Update the cost-per-active-user metric. Report adoption rates to the CIO/CFO.
Annually (at EA renewal or true-up): Determine the right-sized Copilot licence count based on 12 months of usage data. Negotiate licence adjustments (expansion or reduction) with Microsoft. Update the ROI business case with actual data for the next budget cycle. Factor Copilot costs into the EA renewal preparation.
Ongoing: Maintain the role-specific prompt libraries. Update department champion assignments. Monitor for new Copilot features that may unlock adoption in previously low-adoption departments (Microsoft releases Copilot feature updates monthly, and new capabilities can change the adoption equation for specific roles).
“The most expensive Microsoft licence in most enterprises is not the one with the highest per-user price. It is the one that is paid for and never used. At $30/user/month, a Copilot licence that sits idle for a year costs $360 — which is exactly the same as $360 set on fire. The enterprises that extract the most value from Copilot are not the ones with the most licences. They are the ones that assign every licence to a user who will use it, train every user to use it well, and reassign every licence the moment it stops generating returns. Copilot is not a perk. It is a productivity investment that demands the same governance as any other seven-figure line item.” — Fredrik Filipsson, Co-Founder, Redress Compliance
Navigate to the Microsoft 365 admin centre, go to Reports > Usage, and look for the Microsoft 365 Copilot section. You will see aggregate adoption metrics including total enabled users, active users (7-day, 30-day, 180-day windows), and application-level usage breakdowns. You need Global Administrator or Report Reader permissions. For user-level detail, you may need to enable user-level reporting in the admin centre’s privacy settings, subject to your organisation’s privacy policies.
At 90 days post-deployment, a healthy adoption profile is 70%+ of licensed users showing at least one interaction per month, with 50%+ showing weekly usage. Power user rates (daily usage) of 15–25% are typical even in successful deployments. If your 30-day active user rate is below 50%, the deployment has an adoption problem that requires intervention — training, role-specific guidance, or champion programmes.
It depends on the specific terms negotiated in your EA. Some EAs allow mid-term reduction of subscription products like Copilot; others require maintaining the committed quantity until renewal. Check the step-down provisions in your EA. If mid-term reduction is not available, the optimal strategy is to reassign underutilised licences to higher-potential users (maintaining the total count while maximising value) and right-size the count at the next renewal based on usage data.
The most practical ROI model uses time savings: multiply each user tier’s average daily time savings (power users: 30–90 min, occasional users: 5–15 min, non-users: 0 min) by the user’s fully loaded hourly cost, then aggregate across the licensed user base. Compare the total monthly value against the total monthly Copilot cost. A more comprehensive ROI model adds quality improvements (better proposals, fewer errors) and capacity creation (projects enabled by freed-up time), typically measured through structured user surveys.
Across enterprise deployments we have assessed, 35–50% of Copilot licences show minimal or no usage (fewer than 5 interactions per month) at the 90-day mark. This is consistent with adoption patterns for other productivity tools: not every licensed user derives value from every feature. The key is not to accept this rate as permanent — targeted training converts 15–30% of non-users into occasional or power users, and licence reassignment redirects the remainder to higher-value users.
Redress Compliance provides independent Copilot adoption assessments: usage data analysis, ROI modelling, licence optimisation recommendations, and renewal negotiation support based on actual deployment data. We help you pay for the Copilot licences that generate value and stop paying for the ones that do not.