What should you do when adoption is low, in short?

Treat low use as a symptom with a few common causes: people do not know what to use the tool for, are unsure what they are allowed to paste in, tried it once and got a poor answer, or have no time. Each cause has a different fix, so diagnose first.

In most small teams the practical answer is narrow: pick one real, repeatable task per role, show a good example, let people practise it with feedback, and make it normal to talk about what worked. Broad encouragement to "use AI more" rarely changes behaviour on its own.

Measure

Check who has access and who actually uses it, using the admin report your tool provides.

Ask

Spend ten minutes with three or four people asking what stops them.

Clarify

Publish a one-page rule on approved tools and what information is off limits.

Narrow

Choose one weekly task per role where the tool helps and the risk is low.

Practise

Short sessions with a visible standard and a manager who reviews the result.

How do you see who is actually using the tool?

Start with the usage report your vendor provides rather than guessing. Microsoft's documentation for the Microsoft Copilot usage report, dated 14 July 2026, describes enabled users, active users and an active users rate, viewable over 7, 28, 90 or 180 days, plus per-app adoption and a per-user last activity date. It notes that user-specific details are hidden by default.

Use the numbers to find patterns, such as a whole team with no activity, rather than to single out individuals. Activity counts show that someone submitted prompts, not that the output was good, so they are a starting point for a conversation, not a measure of skill.

Why do people not use an AI tool they have been given?

The usual reasons are practical. People cannot see which of their tasks the tool helps with. They worry about breaking a rule because nobody said what is allowed. Their first attempt, often a short vague prompt, produced generic text and they concluded the tool was not useful. Or the tool lives in a place they do not work.

Some hesitation is sensible. A recruiter who is unsure whether candidate CVs may be pasted into a chat tool is right to hold back until someone answers. Fixing that is a policy job, not a motivation job.

Want to see how this connects to a role?

Explore the planned team trial ↗

What does a fix look like in practice?

Brightwell Lettings, a fictional 30-person property agency, licensed Copilot for everyone and saw activity in only one team. Short conversations showed the lettings negotiators did not know whether tenant details could be used, and had tried one-line prompts that produced bland property descriptions.

The office manager published a one-page rule: no tenant personal data in prompts, property facts from the listing sheet only. Negotiators then practised one task: turning a listing sheet into a draft description. A weak prompt read "write a property description". The practised version read "Using only the facts below, write a 120-word description for young professionals. Do not mention anything not listed. Flag any fact you are unsure of." The negotiator then checked every claim against the sheet before use. One concrete, permitted, reviewed task gave the team a reason to open the tool each week.

What adoption help do vendors already publish?

Vendors publish adoption material worth using before you pay for anything else. Microsoft's Copilot adoption hub, checked 16 October 2026, sets out phases from getting ready through to improving, and suggests identifying Copilot champions who help peers, building a community and starting with scenarios tied to team goals. It also offers a prompt gallery and launch templates.

That material is strongest on the tool and the rollout. What a small team usually still needs is practice on its own tasks, with someone checking that the output is accurate and appropriate.

Where does Tosheo.ai fit?

Tosheo.ai is being built for the practice step: role-specific lessons that end with a real task in the tool your team uses, including Copilot and ChatGPT, graded against a visible rubric. Managers are planned to see skill progress and completion, not raw chat logs, and nudges are planned for inactive learners. It does not replace your usage report or your policy.

When is Tosheo.ai not the answer?

If the cause is access, licences or the tool not working inside your systems, that is an IT task. If the cause is an unanswered rule, write the rule first. If you need in-the-flow help inside Teams or a browser extension, those are not included at launch, and vendor adoption programmes may suit you better.

Sources and scope

Microsoft Learn, Microsoft Copilot usage report, 14 July 2026 ↗

Microsoft Adoption, Microsoft Copilot adoption hub, checked 16 October 2026 ↗