Your business doesn't need more AI, it needs better questions

Every week brings a new AI announcement, a new agent, a new pricing tier and a new promise of transformation. It's easy to feel like the only sensible response is to buy something, quickly, before you're left behind. That instinct is the reason so many AI projects underdeliver. When the conversation starts with a product, the shape of the answer is decided before the real problem has been described.

The organisations getting the most out of AI right now aren't the ones with the biggest tooling budgets. They're the ones asking sharper questions about their own business.

Why AI conversations start in the wrong place

Most AI conversations begin with a vendor demo, a headline in the trade press, or a peer's LinkedIn post about a shiny new agent. Those inputs are all useful, but they anchor the conversation to a specific technology before anyone has asked what the business actually needs. From that point on, everything is filtered through "how do we use this thing?" rather than "what problem are we trying to solve?".

The consequence is predictable. You end up with a proof of concept that works, a stakeholder deck that looks polished, and no measurable change in how the business runs. The tool did what it said on the tin. The business just didn't need that tin.

The questions leaders should be asking first

The most useful AI conversations start upstream of technology. Try these instead:

• Where are we losing the most time to work that nobody enjoys?
• Where do decisions get delayed because we can't get to the right information fast enough?
• Where do good employees leave because the tools make their jobs harder than they should be?
• Which processes would we redesign if we were starting the business today?
• What does a great outcome look like, in numbers, six months from now?

Notice how none of these mention AI. They're business questions. That's the point. Once you know which of them matters most, the technology conversation becomes easier because the criteria are already clear.

Finding high-value opportunities inside existing processes

You don't need a workshop with a hundred sticky notes to find AI opportunities. You need to spend an afternoon with the teams doing the work. Sit with your sales operations lead as they prepare a proposal. Sit with your finance team during a month-end close. Sit with your support team as they handle a repeat query for the fifteenth time that week. The friction points will announce themselves.

Once you've seen the friction, ask which of it is repetitive, rules-based and text-heavy. Those characteristics are where AI genuinely earns its keep today. Anything that needs true judgement, context or client relationship still belongs with a human, ideally one who's now got time to do it properly because the boring bits have been handed off.

Common mistakes when evaluating AI solutions

Most poor AI decisions share a handful of root causes:

• Buying to keep up. Adopting a tool because a competitor mentioned it in a webinar isn't strategy, it's anxiety.
• Solving for the wrong user. A tool that thrills the IT team but is ignored by frontline staff has failed, regardless of how good the demo was.
• Ignoring data readiness. AI is only as useful as the information it can reach. If your data is scattered, mislabelled or locked in disconnected systems, expect disappointing answers.
• Treating a pilot as proof. A polished pilot in a lab tells you nothing about how the tool behaves under real workloads, real users and real interruptions.
• No success measure. If nobody agreed at the start what "good" looks like, everyone will disagree at the end about whether it worked.

Measurable outcomes beat impressive demos

The best question to ask before signing off an AI project is a boring one: how will we know this worked? Answers should be specific. "Cut proposal turnaround from five days to two." "Deflect a third of Tier-1 support tickets." "Reduce month-end close from eight days to five." "Free the operations team from the weekly report-building marathon."

Numbers like these force clarity. They also give the people delivering the project something concrete to design against, and give leadership something honest to measure a few months later.

Strategic conversations lead to better technology decisions

When you start with sharp questions, the technology decisions become surprisingly straightforward. You'll usually find that a chunk of what you need is already sitting inside Microsoft 365, that a well-scoped Copilot pilot solves more than a bespoke build would, and that the biggest wins come from tightening a handful of processes rather than adopting a new platform outright.

You'll also find that the internal conversation changes. It stops being about who's champion of which tool, and starts being about what the business is trying to become. That shift is usually where transformation quietly begins.

Where to start

If you're being asked to make an AI call and it doesn't feel grounded, take the technology off the table for an hour. Get the people closest to the work in a room and ask them where the day loses its shape. What you hear will point you at a better first project than any vendor slide deck will. If you want a partner to run that conversation with you, our AI consultancy is built around exactly that starting point: better questions first, better technology second.