Every tool and task category your people have actually tried, with what happened and what to do next. A usage dashboard reports that adoption fell. This reports why, because the reason was captured at the moment someone gave up.
Move this work to Claude Enterprise.
Microsoft 365 Copilot holds 18% adoption across 17 attempts here, while Claude Enterprise holds 92% across 24. You already pay for both.
Do next. Re-route writing and documents to Claude Enterprise. No procurement needed — this is a routing change, not a purchase.
Stop routing this here — "It was faster to just do it myself".
40% adoption across 15 attempts, and 56% of rejections cite "It was faster to just do it myself" — a property of the task, not the prompt.
Do next. Negative time-to-value. The task is too small or too bespoke to carry the cost of instructing a model. Retire it from the playbook and redirect the licence attention to a higher-frequency task.
Fixable — right tool, wrong execution.
50% adoption across 10 attempts, 60% of rejections citing "It didn't fit how I actually work".
Do next. Routing problem. The task suits AI but landed in the wrong place — the tool sits beside the work instead of inside it. Re-route to a tool that reaches the system where the work already happens, or make it an agent that runs in that process.
Fixable — right tool, wrong execution.
57% adoption across 7 attempts, 67% of rejections citing "It didn't fit how I actually work".
Do next. Routing problem. The task suits AI but landed in the wrong place — the tool sits beside the work instead of inside it. Re-route to a tool that reaches the system where the work already happens, or make it an agent that runs in that process.
6 tool and category pairs have fewer than five recorded attempts. Adopt will not rule on those — a verdict from three data points is a guess wearing a badge.