Adoption Isn't the Same as Trust

When Copilot is the only AI tool available to an employee, adoption reaches 68 percent, according to Recon Analytics research published this year. Give that same employee a free choice between Copilot, ChatGPT, and Gemini, and usage falls to 8 percent.

What the usage number actually measures

A dashboard showing 68 percent adoption looks like success from the outside. It measures compliance more than preference. An employee with one tool and a mandate to use it will show up in that number whether the tool works well for them or not. Take away the mandate and give them alternatives, and the real number, 8 percent, is what's left once choice enters the picture.

Every executive dashboard I've reviewed stops at usage: rollout complete, licenses activated, seats provisioned, adoption declared a success. The harder question almost never gets asked: would these same people keep using it if they had any other option?

Why teams don't say so out loud

Nobody schedules a meeting to announce they've stopped trusting a mandated tool. There's no field in the adoption dashboard for it, and flagging distrust of a leadership-backed rollout carries a small but real professional cost. Recon Analytics found that 44 percent of people who tried Copilot and later quit named the same reason above all others: they stopped trusting what it told them. The people who stay have usually just found it easier to comply than to explain why they wouldn't.

What checked-out adoption looks like from the inside

A team can be fully adopted and quietly checked out at the same time. They click through the tool because it's mandatory, they don't raise it with leadership, and they've privately stopped believing what it produces. The tells are usually small: output copied without review, prompts kept minimal because nobody expects the result to be worth refining, real verification work happening somewhere the dashboard doesn't see. On a usage dashboard, all of that looks exactly like success.

A better measurement

Ask a different question in the next adoption review: how many would choose this if they didn't have to. It's a harder number to get and a much more honest one.

The fast layer will keep shipping new features regardless. Whether anyone still trusts what comes out of them is a much slower thing to earn back.