Every AI project I’ve seen go wrong has had the same root cause. Not the wrong tool. Not the wrong vendor. Not even a bad team. The problem was that nobody asked — really asked — what they were trying to fix.
So before anything else, I ask one question: What decision will be better, or what task will be faster, if this works?
That’s it. Simple. But most teams can’t answer it clearly, and that’s the whole problem right there.
Why this question matters
AI is good at specific things. It can classify, summarise, draft, predict, and retrieve at speed and scale. What it cannot do is compensate for a brief that doesn’t know what it wants.
When I work with a business that wants to “use AI,” the first conversation is rarely about technology. It’s about operations. What’s slow? What’s error-prone? Where do people spend time on work that feels mechanical — copying data, summarising the same things, routing the same requests?
The answer to that question is your brief. Everything else — the tool choice, the integration, the rollout — follows from it.
The trap most businesses fall into
The most common failure mode I see is what I call the solution-first problem. A senior leader sees a demo, gets excited, and the team is tasked with “implementing AI.” Nobody says what problem they’re solving. The team picks a tool. The tool gets deployed. Six months later, adoption is poor and nobody’s quite sure why.
The honest answer is usually that the tool was solving a problem the business didn’t actually have, or had but didn’t feel acutely enough to change behaviour for.
A more useful way to start
Instead of “how can we use AI?”, try these questions:
What takes too long? List the ten tasks your team does most often that feel like they shouldn’t take as long as they do. Not the complex ones — the mechanical ones.
What gets done inconsistently? Inconsistency is a signal that the task relies too heavily on individual judgement, memory, or effort. That’s often exactly where a well-designed AI system helps.
What doesn’t happen because there isn’t time? This is the quieter version of the same problem. These are the tasks that fall off the list when people are busy — analysis, follow-up, review. AI can pick those up.
What you do with the answers
Once you have honest answers to those questions, you have a real brief. Not “we want to use AI,” but “we want to cut the time it takes to produce monthly ops reports from three days to three hours” or “we want every customer enquiry triaged and summarised before it reaches a human.”
Those are briefs you can build to. You can measure progress, you can define done, and you can tell whether it worked.
That’s when we start talking about tools.
If you’re at the “where do we even start?” stage, a strategy call is the fastest way to get from vague interest to a clear first step.