Executive answer
The problem: Microsoft’s public tools can identify licensed employees who have not used Copilot in 30 days, send them a reminder, and measure views and clicks. But inactivity says nothing about why a person has not used it.
The recommendation: Add one optional, role-aware diagnostic question before or alongside the reminder. Then route people to the right next step: a relevant use case, clearer policy, practical support—or no prompt at all.
The decision consequence: Without that small diagnostic step, an organization may spend more effort persuading employees when it should be removing a barrier.
Independent public-material review. This article examines Microsoft documentation available to the public. Violet was not commissioned by Microsoft, has no affiliation with Microsoft, and has not observed a customer deployment or measured the effect of the tools discussed here. Documented features and Violet’s hypotheses are identified separately.
The public workflow in one minute
Microsoft’s public guidance lets administrators identify licensed users who have not used Copilot during a rolling 30-day window. They can send a preset in-product message and review messages seen, clicks and click-through rate. (Microsoft Learn: Organizational Messaging)
The path is roughly:
License assigned → no recorded use → reminder → click → learning resource
This is efficient, and it can solve a real problem: an employee has a suitable use case but does not know Copilot is available or needs a simple starting point.
The difficulty is that the system sees absence. It cannot see the reason for the absence—and that reason changes the intervention.
One signal can hide six different states
Two employees can look identical in a usage report and require opposite responses.
| What the dashboard sees | What may actually be happening | What to test next |
|---|---|---|
| No use in 30 days | “I did not know I had access.” | Awareness and access |
| No use in 30 days | “I cannot see a useful task for my role.” | Task fit before promotion |
| No use in 30 days | “I am unsure what data or tasks are allowed.” | Policy clarity and escalation routes |
| No use in 30 days | “I tried it and did not trust the result.” | Output quality and verification burden |
| No use in 30 days | “The training did not help when I needed it.” | Practice on a real, role-specific task |
| No use in 30 days | “It is not worth using for this task.” | Whether non-use is a sensible choice |
A dashboard cannot establish which explanation applies. A reminder may help the first employee, frustrate the second, and create risk for the third.
Before choosing the words, the team needs one more bit of information.
The recommended intervention
Do not replace a small reminder with a large survey. Add one lightweight, optional fork:
Which best describes what is getting in the way right now?
- I have not found a task where Copilot would help.
- I am unsure what I am allowed to use it for.
- I need an example for my role.
- I tried it, but the result was not useful enough.
- I do not have time to learn it right now.
- None of these.
The answer determines the next message.
| Reported barrier | More useful next step |
|---|---|
| No relevant task | Show one recurring, role-specific task—and allow “not relevant” |
| Policy uncertainty | Give plain examples of permitted and prohibited use, plus a named contact |
| Need an example | Demonstrate a short workflow instead of promising generic productivity |
| Weak result | Help with task selection, prompting and verification; acknowledge limitations |
| No time | Offer a low-cost practice step when a relevant task appears |
This is still a hypothesis. Employees may ignore the question, the choices may be too coarse, or a generic reminder may already work for most people. The diagnostic fork earns its place only if it improves what happens next.
But it changes the starting assumption. A reminder can now address a named problem instead of standing in for a diagnosis.
Why this is a narrow critique
It would be unfair to treat one notification feature as Microsoft’s entire adoption theory.
Microsoft’s wider rollout guidance recommends setting goals, use cases and success measures; starting with a limited group; gathering feedback; tailoring training; and refining the rollout. Its adoption guide also recommends executive sponsorship, champions, learning communities, surveys and periodic business reviews. (Microsoft Learn: Roll out Microsoft 365 Copilot, Microsoft Adoption: Essential guide)
Those practices can distinguish an awareness problem from a workflow, trust, skill or policy problem. The opportunity is to connect that richer adoption practice to the moment an administrator turns low-use data into an intervention.
Measure adoption beyond activity
Views and clicks reveal exposure and immediate engagement. They do not establish that a person found a suitable task, used the tool safely, improved the work or wanted to return.
Microsoft’s AI adoption score uses active days as an operational measure, with a stated target of using Copilot on an average of three days each week. That can be useful, but activity cannot answer every adoption question. (Microsoft Learn: AI adoption score)
A stronger test separates:
- Exposure: Did the intended group encounter the message?
- Barrier: What prevented the next useful action?
- Task fit: Did the person identify a relevant, permitted task?
- Capability: Could they complete it with appropriate verification?
- Value: Did the result improve a declared work outcome?
- Guardrails: Did confusion, rework, risky use or pressure increase?
- Choice: Did some employees make an informed decision not to use it?
If the diagnostic fork does not improve those downstream measures, it should not survive.
What would prove Violet wrong?
This review does not claim that reminders are ineffective. It claims that a reminder addresses a particular kind of barrier and should not become the default explanation for every inactive user.
The hypothesis would be weakened if generic reminders consistently improved relevant task completion, confidence, work quality and sustained use across roles—without meaningful differences in trust, task fit or policy uncertainty.
It would also be weakened if the diagnostic question added friction without improving downstream results. Either finding would support the simpler path.
The aim is not to preserve complexity. It is to keep only the complexity that changes the next decision.
The decision behind the reminder
People who have not used a tool may need a reminder. They may instead need a relevant task, a clearer policy, a safer workflow, better practice, more credible evidence—or permission not to use the tool where it does not help.
So the useful executive question is not:
How do we make inactive users use AI?
It is:
What would make the next use relevant, safe, informed and worth repeating for this employee in this role?
For an AI rollout, launch message or low-adoption journey, Violet can help identify the first broken transition and design a bounded test. Start a focused message audit or discuss a larger AI-adoption journey.