Monitoring AI usage is valuable — as long as it does not quietly become a proxy for performance.
A trend is beginning to take shape. Organisations are starting to track how much employees are using AI in their day-to-day work. Employees are being asked to show where they are using AI — not necessarily what outcomes it is driving. Enterprise-wide AI adoption is being pushed as a measurable metric.
On the face of it, this makes sense. Every technology shift requires behaviour change. If people do not use AI, the value will never be realised. So naturally, leaders are pushing for adoption.
But I keep coming back to a more fundamental question.
Are we optimising for AI usage — or for enterprise outcomes enabled by AI?
Because the two are not the same. AI usage is an input metric.
Enterprise value is an outcome metric. And when input metrics become the goal, organisations often end up optimising for visibility rather than impact.
The recent vulnerability reported in McKinsey’s internal AI tool is one example. Broader observations from firms like Deloitte also point to governance and oversight struggling to keep pace with AI-led decisioning. These are reminders of a deeper truth:
When adoption runs ahead of architecture, systems become fragile.
As we push adoption, we need to anchor ourselves in a few first principles:
- Are we designing a control plane for AI, or stitching together disconnected tools, copilots, and scripts?
- Are we explicit about decision rights — where AI assists, where it recommends, and where it is allowed to act autonomously?
- Do we have observability and auditability — can we trace how decisions are made, by which agents, on what data?
- Are we treating AI as a new attack surface with defined trust boundaries — or as just another productivity layer?
- Are we measuring activity (prompts, usage, adoption) or outcomes (cycle time, quality, resilience, business impact)?
As we move from AI as a tool to AI as a participant in enterprise systems, what enterprises need is architecture, governance, and accountability — not just adoption.