Two-Thirds of Copilot Licences Go Unused. Check Your AI Spend.

Microsoft sells Copilot at $30 per user per month. The pitch: 14 minutes saved per employee per day, embedded directly in Word, Excel, Outlook, and Teams. For a 15-person accounting firm, that’s $5,400 per year. Sounds reasonable — until you look at who actually opens it.

According to data tracked by Recon Analytics through January 2026, 64% of Microsoft 365 Copilot seats are dormant. Not occasionally idle. Dormant. Two in three employees with paid access don’t use it in a given month. At $30 per seat, that’s roughly $3,456 per year in dead spend for that 15-person firm — money spent on a tool that sits in the toolbar like an unused gym membership.

The activation gap is only part of the story. Recon Analytics tracked Copilot’s accuracy Net Promoter Score across three points: -3.5 in July 2025, crashing to -24.1 in September, and recovering slightly to -19.8 in January 2026. A negative NPS means more detractors than promoters. Among users who stopped using Copilot entirely, 44.2% cited distrust of its answers as the primary reason.

The preference data is more telling. When employees have access to both Copilot and ChatGPT, 76% choose ChatGPT. When all three major platforms — ChatGPT, Gemini, and Copilot — are available with equal friction, Copilot’s share falls to 8%. But when Copilot is the only tool available, adoption reaches 68%. That gap reveals something important: most Copilot usage is driven by employer mandate, not employee choice.

Copilot is a capable tool used badly by most organisations. The pattern repeats across every AI purchase: a business buys seats for the whole team, sends a “Copilot is now available” email, and waits for productivity to appear. It doesn’t.

Business.com’s 2026 Small Business AI Outlook found that while 57% of US small businesses now invest in AI, only 30% of employees use it daily. Seven percent of businesses publicly endorse AI without implementing it at all. We’ve written before about the gap between AI adoption and AI value among Australian SMEs — only 5% report getting real value from their tools, per Deloitte. The pattern is the same: buying is easy, embedding is hard.

The businesses getting value — the ones reporting 5.6 hours saved per employee per week, per Business.com — share one trait. They started with a specific workflow, not a general tool. They identified that document drafting was taking three hours, or that data entry consumed every Monday morning, or that client follow-up emails were falling through the cracks. Then they matched a tool to that workflow, trained the people doing it, and measured whether it stuck.

Accounting firms and law practices are the primary target market for Microsoft 365 Copilot. They’re heavy Microsoft 365 users, their work is document-intensive, and the AI integration sounds like a natural fit. Many are already being pitched Copilot by their IT providers or Microsoft partners.

Before signing up, ask one question: what specific task will each person use this for? If the answer is vague — “it’ll help with emails” or “it should speed things up generally” — the data says you’ll join the 64%. The firms seeing returns are more specific. A tax team using Copilot to draft client correspondence from working papers. A paralegal using it to summarise long contracts. A bookkeeper using it to reconcile data across spreadsheets. Each use case is narrow, trainable, and measurable. The rest of the firm doesn’t need a licence.

This is fundamentally a cost intelligence question. The cost of AI isn’t the subscription. It’s the subscription multiplied by the gap between how many seats you bought and how many actually work.

Audit what you’re already paying for. Check your Microsoft 365 admin centre for Copilot usage reports — they’re built in. Check ChatGPT Team or Claude for Business dashboards if you have those. For any tool with less than 40% active usage, you have three options: train the non-users on a specific workflow, reassign the licences to people who will use them, or cancel.

Then apply the same test before buying anything new. Name the task. Name the person. Set a 30-day check. If you can’t do all three, you’re buying a gym membership.

Key takeaways

Not sure which AI tools are worth keeping?

A short conversation can help you audit what you’re paying for, identify the workflows worth automating, and avoid the 64% trap. Book a call to talk it through.