Organisations increasingly have access to detailed AI usage data: licences, active users, training completion, frequency of use and feature activity. These are valuable signals. But usage alone does not tell us whether AI has become part of everyday work or whether the work itself has changed. In this practical 20-minute session, Selina Thompson explores the missing link between AI activity and evidence of organisational adoption. Drawing on Tracked, the measurement element of her TRUST framework for AI adoption, she introduces five layers of evidence: access, usage, behaviour, workflow and outcome. These are not presented as a causal chain, but as different sources of evidence that help organisations understand what is changing around AI-enabled work. Attendees will leave with four practical questions they can take back to their organisations: - Which roles and workflows are we expecting AI to change? - What behaviour would indicate adoption, and where would it leave evidence? - What would be different in the workflow if the new way of working were becoming embedded? - Which workforce or business outcome should improve, what must not get worse, and over what timeframe? Rather than proposing another universal set of AI KPIs, the session shows how organisations can identify the evidence that matters for the roles, workflows and outcomes they are trying to change. The session is designed for data, AI, digital transformation and change leaders who need to understand whether AI is moving beyond deployment and experimentation towards sustained changes in how work gets done.