What Microsoft’s rollout actually showed
Following Cowork’s internal debut in April 2026, Microsoft combined practical guidance, recurring promotion and a June employee-story campaign, #HowICowork. The company reports these results:
| Measure | Reported result |
|---|---|
| Campaign reach | About 58% of employees |
| Employee contributions | 45 people shared prompts and stories |
| Weekly active users during the campaign comparison | About 35,000 to nearly 45,000, a reported 27% increase |
| Monthly active users, mid-August to early October | About 60,000 to 70,900 |
| Early-October activity | 52,400 weekly and 27,800 daily active users |
Usage was already growing before the campaign, and Microsoft lacked a clean pre-launch baseline. These are company-reported adoption figures, not a controlled measurement of campaign effectiveness.
That distinction matters. A rising activity chart can support a decision to investigate what worked; it cannot, by itself, tell an IT department which intervention deserves the credit—or whether employees completed better work.
There is also a product-timeline boundary worth preserving. Microsoft Learn records Cowork’s general availability for Microsoft 365 Copilot tenants worldwide in tier-1 languages in June 2026. Microsoft’s April employee debut was an internal milestone, not the same event as general customer availability.
Define “active” before celebrating adoption
Microsoft’s administrator documentation adds an important qualification: an active Cowork user can be someone who submitted a prompt or someone on whose behalf a scheduled task executed. The retrospective does not establish that its internal figures use exactly this reporting methodology. Nevertheless, administrators should understand the distinction before interpreting their own numbers.
Automated activity is not inherently less valuable. A scheduled task that reliably delivers useful work may represent successful delegation. But it answers a different question from “How many employees deliberately returned to Cowork this week?”
For an authorized administrator, Microsoft documents this reporting path:
- Sign in to the Microsoft 365 admin center.
- Select Copilot > Cowork.
- Open Usage and choose a reporting date range.
- Check the report’s refresh timestamp.
- Compare user-initiated and scheduled tasks, returning users and task volumes; export the details when needed.
The default reporting window is 28 days. Microsoft defines retained users as people active in both the previous and most recent seven-day periods.
One further caution: the Frontier-preview Value tab’s “assisted hours” are estimates—not measured time savings, productivity gains or realized business value. Treat them as directional indicators, not payroll-ready arithmetic.
A practical framework for enterprise IT
Microsoft’s experience suggests a sensible rollout framework. The following is an operational interpretation, not a claim that Microsoft experimentally validated each step.
Start with a recognizable task
Ask participating employees to document a bounded piece of work: the intended outcome, the context supplied, the result and the corrections required.
A useful example should help another employee decide whether to try the workflow. A polished output without the review effort tells only half the story.
Establish measurement before promotion
Record an initial usage period where possible, then mark campaign dates, training sessions, access changes and significant product updates. Keep reporting windows consistent.
The goal is not to turn every internal announcement into a laboratory experiment. It is to avoid discovering, months later, that nobody can distinguish increased access from stronger retention—or either from a seasonal change in workload.
Make feedback actionable
Give employees a clear place to report unsuccessful tasks and unclear guidance. Classify the problem before choosing the response:
- Understanding: The employee does not know what to delegate.
- Access or configuration: The intended workflow is unavailable.
- Output quality: The result requires excessive correction.
- Workflow fit: The task is technically possible but not worth delegating.
These categories are proposed triage tools, not findings from Microsoft’s published campaign data. Their purpose is to prevent every complaint from receiving the same answer: another training email.
Examples, integrations and discoverable help
Microsoft also built plugins, including Azure DevOps and Chart, reused employee examples in a seller-facing campaign kit, and reported that guidance continued helping through search and Copilot after promotion ended. The retrospective does not establish external availability for those internal plugins or quantify customer outcomes from the kit.
For other organizations, the practical question is whether guidance remains useful at the moment of need. Consider testing whether employees can find an accurate answer to a common workflow question—not merely whether they opened the launch newsletter.
That is a proposed evaluation method, rather than evidence that Copilot consistently retrieves the right guidance. Keep page traffic, answer accuracy and successful task completion separate.
The next challenge is sustained delegation
Microsoft’s stated next steps include bringing Chat and Cowork together in Copilot Home, eventual automatic routing, and an “Autopilot” shift toward ongoing delegation. The retrospective supplies no delivery dates for those future changes.
The strongest lesson is therefore not a promise about the next AI experience. It is a management discipline: make useful work visible, measure repeated engagement carefully, and investigate whether the results justify the effort.
Launch day can establish availability. A successful rollout still needs evidence that people can—and want to—get worthwhile work done.