AI adoption did not wait for your steering committee. The more useful question is whether anyone knows which tools people are using, what they are putting into them and what comes back out.
Buying the software is easy. Cleaning up permissions, changing working habits, finding useful workflows and proving anyone is actually getting value from it is the expensive bit.
The strongest AI strategies are usually less obsessed with AI. They start with revenue, cost, customers, friction and competitive advantage, then ask where AI genuinely deserves a seat at the table.
Asking AI for an answer is one thing. Letting it send the email, update the system, move the money or make the decision is where the conversation gets considerably more interesting.
Governance is not measured by the number of people invited to a monthly meeting. It is whether someone can make a decision, explain it and show what happened next.
A good pilot proves something can work under friendly conditions. Scaling it means dealing with everyone, everything and every awkward exception that was conveniently missing from the demo.
Opening an AI tool is not a business outcome. Neither is writing a prompt. Ten thousand active users can still add up to remarkably little measurable value.
The AI project that saves twelve minutes from a task nobody enjoys may never make the keynote presentation. Do it thousands of times a month and the finance team may start paying attention.
A beautifully formatted policy nobody remembers is not much of a control. The real test is whether people know what to do when the policy collides with an actual deadline.
Watching somebody write five clever prompts does not create organisational capability. People need to know where AI fits into their actual job, and when not to trust it.
Not every AI system arrives with a launch event. Some quietly appear inside software updates, supplier platforms and ordinary workflows before anyone thinks to put them on a register.
Unclear ownership, bad data, broken processes and confused priorities did not arrive with generative AI. AI has simply become remarkably good at exposing them.
