We bought Copilot licences and nobody uses them: how to fix it

To fix Copilot adoption, stop training everyone on everything. Pick a few roles, give each role one real workflow where Copilot does a named step, measure usage before and after, and use trainers who have done that work themselves. Licences sit unused because nobody showed people where the tool fits into the job they already do.

#Key takeaways

• Low Copilot usage is usually a workflow problem, not a licence or feature problem.
• Copilot waits to be asked, so people need a specific moment in their day when asking makes sense.
• Start with a few roles and one workflow each, not a company-wide launch email.
• Measure usage per role before training and again after, so the renewal decision rests on numbers.
• If executives do not use it visibly, staff read that as permission to ignore it.

#Why does nobody use the Copilot licences we bought?

The most common reason buyers give is simple: it does not insert itself into the workflow, it waits to be asked. A finance analyst closing the month or a coordinator chasing approvals does not stop to wonder what Copilot could do. Without a habit attached to a real task, the button stays unclicked and the licence goes unused.

The second reason is that people already have a tool they like. Many staff still use ChatGPT on their own, often with company data. The Microsoft and LinkedIn Work Trend Index, may 2024 found that 78% of AI users bring their own AI tools to work, and only 39% of AI users had received AI training from their company.

The third reason sits at the top. When executives do not use Copilot themselves, teams conclude it is optional. Adoption follows what leaders do in meetings, not what the launch email said.

#What actually raises Copilot usage?

Treat it as a set of small workflow changes, not a rollout. Stennir runs it in this order:

1. Pick a few roles where the work is repetitive and written: finance, HR, procurement, customer service leads.
2. For each role, choose one workflow and name the exact step Copilot handles, such as drafting the variance commentary or summarising a supplier thread.
3. Record current usage per role from the admin reports before any training starts.
4. Train each role as its own cohort, on its own documents, with a trainer who has done that job.
5. Measure usage again and compare, role by role.

The order matters. A short session on a real task beats a long generic course, and pairing confident users with hesitant ones in the same cohort spreads the habit faster than any slide deck.

#How do we decide what to do before the licences renew?

Renewal is the moment buyers worry most, because nobody can say whether the licences paid off. That is why the before-and-after measurement matters. If a role's usage rises once it has a workflow, keep those licences. If a role still does not use the tool, reassign or drop them, rather than renewing on hope.

This is not unusual. Gartner, july 2024 predicted that 30% of generative AI projects would be abandoned after proof of concept, citing unclear business value among the causes. An adoption report per role gives you a clear answer instead of an impression.

#Who should run the training?

Practitioners, not presenters. People trust a trainer who has closed the books or run a support queue and can show the prompt they actually use. A generic trainer can explain features; a practitioner can show where the feature saves a step on Monday morning.

This is how the AI training service is built: the AI Operator Program runs role-specific cohorts on the real workflow and ends with an adoption report. For the operations view, see AI for operations.

#What should an HR or it lead do next?

Pull the usage report, list the roles with the lowest numbers, and pick the one where a single workflow would make the most difference. If you want an outside view first, take the free readiness assessment or book a 30-minute call. Our guide to choosing an AI training company covers what to ask any provider.

#Common questions