The most common question leadership teams ask when they get serious about AI isn't "what's possible?" It's "where do we actually begin?" And it's the right question to be asking — because starting in the wrong place is exactly how most AI initiatives fail.
Most businesses we work with have no shortage of AI ideas. The problem isn't imagination — it's prioritisation. Without a clear framework for deciding where to start, organisations scatter their energy across disconnected pilots, run out of momentum, and conclude that AI "just doesn't work for us."
It does work. But it works when you start in the right place, with the right problem, and with realistic expectations about what it takes to get there.
Why Most Companies Start Wrong
There are three failure modes we see consistently across industries:
They start with the technology, not the business problem. "We want to use AI" is not a strategy. "We want to reduce unplanned downtime by 20% in our top facility" is a problem worth solving — and AI may be the right tool to solve it.
They start with hype, not outcomes. The pressure to "do something with AI" leads to chatbot deployments and internal tools that nobody uses. These feel like progress but deliver no measurable return.
They start everywhere at once. Without prioritisation, resources spread thin. Nothing gets the focus it needs to succeed, and leadership loses confidence in the whole programme.
Three Questions to Ask Before Any AI Investment
Before committing budget or team time to any AI initiative, every leadership team should be able to answer these three questions clearly:
1. What business problem are we actually solving?
Be specific. Not "improve customer experience" but "reduce average call handling time by 15% without increasing headcount." The more precisely you can define the problem, the easier it is to evaluate whether AI is the right solution — and to measure whether it worked.
2. Do we have the data to solve it?
AI runs on data. Before investing in any AI initiative, audit what data you have, where it lives, how clean it is, and whether you have the volume required. Many AI projects stall not because the model doesn't work — but because the underlying data was never ready. A data readiness assessment should be the first step of any AI programme.
3. Do we have the organisational readiness to act on the output?
This is the question most organisations skip — and the one that kills more AI projects than any technical problem. If your AI system identifies a maintenance issue 48 hours before failure, does your operations team have a process to act on that alert? If your demand forecasting model updates every night, does procurement know how to use it? AI creates value when organisations are ready to act on what it tells them.
A Simple Framework for Prioritisation
Once you've gathered your list of potential AI use cases, plot them on two dimensions: business impact and feasibility. Business impact is the value delivered if the initiative succeeds — revenue growth, cost reduction, risk mitigation, competitive advantage. Feasibility is a combined score of data readiness, technical complexity, and organisational readiness.
The top-right quadrant — high impact, high feasibility — is where you start. These are your quick wins: initiatives that build momentum, demonstrate ROI, and create organisational confidence in AI. The top-left quadrant (high impact, lower feasibility) is your strategic roadmap — the bigger bets that require foundational work before they can succeed.
The goal of the first AI initiative is not to solve the biggest problem in the business. It's to prove that AI works in your organisation, build the muscle to do it again, and earn the confidence to take on bigger challenges.
What Good Looks Like
Organisations that get their first AI initiative right share a few common characteristics:
Clear ownership: one senior sponsor, one delivery lead, clear accountability for outcomes.
Measurable outcomes defined before work begins — not evaluated afterwards.
A narrow, focused scope that can demonstrate value within 90 days.
A roadmap beyond the first initiative, so momentum doesn't stop when the pilot ends.
Executive communication built in — so the board understands what's happening and why.
Getting started is the hardest part. But the businesses that start right — with a clear problem, a realistic assessment of readiness, and a focused first bet — are the ones that build genuine AI capability over time. The ones that start with the technology and hope the strategy follows rarely get there.