The Two Pulls
Leadership wants visibility on AI adoption. Employees fear surveillance. Both concerns are legitimate. The framework below balances them: measure what matters, transparently, without invasive monitoring.
The Right Metrics
- Number of workflows where AI is in regular use, by team.
- Outcomes on those workflows (time saved, errors reduced, output quality).
- Self-reported AI usage in monthly surveys.
- Aggregate spend on AI tools (a proxy for usage).
- Internal champion engagement (who's helping others adopt).
Wrong Metrics
- Per-employee prompt counts.
- Screen recording of AI tool usage.
- Keystroke logging.
- Anything that looks at individual employee AI behavior.
The wrong metrics produce the wrong incentives: employees who feel watched stop experimenting. Adoption craters precisely because you're measuring it.
Self-Reporting
Monthly anonymous survey: where did AI help this month? Where did it hurt? What's missing? Aggregate the answers. Use them for product decisions, not for performance management.
- • Workflows in production per team.
- • Quarterly outcome metrics (vs baseline).
- • Monthly satisfaction survey.
- • Champion network engagement.
- • Tool spend efficiency.
Outcome-Based Tracking
Track outcomes per workflow, not behaviors per person. If support resolution time dropped 60%, that's the metric. Whether agent X used AI 47 times or 4,700 times doesn't matter.
Transparency Builds Trust
Share what you're tracking and why. Publish the survey results back to the team. Demonstrate that the data informs product, not performance review. Trust compounds.
Culture Around Adoption
The cultural moves that drive adoption: celebrating wins openly, creating space to share AI experiments that didn't work, leadership using AI visibly, removing punitive metrics. Adoption follows culture.
Surveillance kills adoption. Visibility helps it. The difference is whether the measurement is at the individual level (don't) or the workflow level (do).
See AI change management.
FAQ
What if someone refuses to use AI? Fine, as long as the work gets done. Forced adoption fails.
Should adoption be in performance reviews? No. Outcomes can be; behaviors should not.
Who owns adoption metrics? The AI champion, reporting to the CEO.