Adoption, not capability, is where AI fails
Most AI tools do not fail because the model is weak. They fail because people get access and never change how they work. The login goes out, the demo happens, and a month later everyone is back in their inbox. Fixing that is the whole job.
How I run it
It starts with the leaders. I run an executive workshop so the leadership team uses Claude on real work and sets the rules. Then it goes department by department, sales, operations, finance, each team learning on the tasks that eat their week, with the first Skills built live so people see their own work get done. The cautious people become the guardrails experts, not the holdouts. The measure is not who attended. It is whose daily work actually changed.
What the research says
The gains are real when adoption is real, and they are largest for the people who need them most. Customer-support agents resolved 14% more issues per hour on average, and 34% more as newcomers (Brynjolfsson, Li & Raymond, QJE 2025). Knowledge workers completed 12% more tasks at over 40% higher quality on work inside AI's strengths (Dell'Acqua et al., Harvard Business School and BCG, 2023). And leaders consistently underestimate how much their own teams already use AI.
Proof
A nonprofit cut a grant application from eight hours to three after a single ninety-minute session.