Where AI creates leverage — and where it doesn’t
The useful question is not “what can AI do?” but “which decision or workflow becomes meaningfully cheaper, faster or better?”
Most organisations now feel some version of the same pressure: we need to do something with AI. The board asks about it. Competitors announce pilots. Vendors add “AI-powered” to products that haven’t changed.
So companies start with the technology and go looking for a problem. That’s backwards, and it’s expensive.
AI without a business case is expensive theatre.
Where it tends to pay
- High-volume, repetitive work with a clear definition of “good enough” — drafting, sorting, summarising, first-line answers
- Knowledge that’s stuck in people’s heads and slows everyone else down when those people are busy
- Preparation for decisions: pulling the facts together faster so humans spend their time on judgement
- Small, specific tools built for one workflow, owned by the people who use them
Where it usually doesn’t
- Processes nobody agrees on yet — automating confusion just makes it faster
- Problems that are really about ownership, incentives or unmade decisions
- “Platforms” bought before a single use case has proven its value
- Anything where the cost of a wrong answer is high and nobody is checking
A simple test
Before any AI initiative, answer three questions in plain language:
What gets better, for whom, and how would we know? If the answer needs a slide full of buzzwords, it isn’t ready. If it fits in one sentence with a number in it, build a small prototype and find out.
The goal isn’t to use AI. The goal is a better business. Sometimes that means building an agent. Sometimes it means fixing the process first. Occasionally it means doing nothing — which is also a decision.