Most companies are now past the point of wondering whether AI matters. The more useful question is what, exactly, they are supposed to do with it.
That sounds obvious, but it is where a lot of teams are currently getting stuck. They have given people access to ChatGPT, Claude, Copilot or Gemini. A few people are using it heavily. A few are dabbling. A few are avoiding it entirely. Somewhere there is probably a Slack channel full of prompts and slightly breathless automations someone built on a Friday afternoon. None of that is a strategy.
The mistake I see is leadership treating AI adoption as a volume problem. More usage, more experiments, more tools. The assumption is that if enough people start using AI, the organisation will somehow become more efficient. It won’t. Not like that.
AI isn’t a silver bullet. It’s a precision instrument. And right now, most teams are using it like a hammer looking for nails.
The real problem isn’t tools. It’s workflows.
Product organisations aren’t slow because people can’t write fast enough. They’re slow because work moves through a messy chain of handoffs, half-remembered decisions, scattered evidence and repeated explanations. Customer feedback lives in support tickets, app reviews, sales calls, research notes, Slack threads and dashboards, simultaneously everywhere and nowhere useful. Research gets presented once, then disappears into a deck nobody opens again.
This is the unglamorous middle layer of product work. Not the vision deck, not the launch campaign, not the beautiful Figma file. The operational glue between teams. And it’s where product organisations quietly haemorrhage time, week after week.
A real example.
I recently built a simple NPS digest using an edge function. Nothing exotic. The inputs were already there, NPS responses coming in regularly, themes repeating, signal getting lost in the noise. What I built classified the feedback, grouped themes, tracked week-on-week changes and surfaced direct customer language. A short, reviewed summary lands on a regular cadence. The team can actually see what’s shifting, what’s persistent, what needs a decision.
It didn’t replace anyone’s judgement. It gave us a better evidence base to exercise judgement against.
That distinction matters. AI shouldn’t decide your product strategy. It should reduce the manual drag between evidence and decision. The human still interprets, prioritises and owns the outcome.
Start with a workflow audit, not a tool.
Ask your team where time and context are repeatedly lost. Where do people copy information between systems? Where do you manually reconstruct the same story every week? Where does customer insight disappear? Where are decisions being made on anecdote because the structured evidence is too hard to retrieve?
Start with the problems you’re facing and you’ll get more success than wasting time asking the team to ‘use AI’.