Deciding where to start with AI at a mid-size company

Illustrative case

This is an illustrative composite, not a real client engagement. The company, the numbers, and the recommendations are invented to show how a typical AI-strategy engagement unfolds and what it produces.

A mid-size services company felt the pressure every leadership team now feels: the board wanted to know "what we're doing about AI," competitors were making announcements, and internally a dozen ideas were floating around — a chatbot, document automation, a predictive tool, something with the sales data. The risk wasn't inaction. It was the opposite: spreading money and attention across too many half-considered projects, or making one expensive bet on the flashiest idea rather than the wisest one.

The real problem

The team didn't need more ideas; they needed a way to choose among the ones they had. So the engagement wasn't about inventing a grand AI vision. It was about bringing discipline to a decision — replacing "this sounds exciting" with "this is worth doing first, and here's why."

How we approached it

We ran a structured assessment built on a simple principle: every candidate use-case gets scored on the same three things, so they can be compared honestly. The three are value (how much it would help if it worked), feasibility (whether the company's data, systems, and skills can actually support it), and risk (what could go wrong — regulatory, reputational, operational). We gathered the candidate ideas, interviewed the people who'd own them, and scored each one. The output isn't a verdict handed down from outside; it's a shared, transparent ranking the team can see the reasoning behind.

What emerged

In this illustrative case, the assessment surfaced about a dozen use-cases. The flashiest — a customer-facing AI assistant — scored high on value but low on feasibility and high on risk, so it landed as "not yet," to be revisited once groundwork was in place. Two unglamorous candidates rose to the top instead: automating a slow internal document process (high feasibility, low risk, solid value) and a churn-prediction tool that built naturally on data the company already had. Several ideas scored low enough across the board to be set aside, which freed up attention as much as the top picks did.

The honest part

The most useful thing we told this team was that their most exciting idea wasn't their best first idea. Strategy work earns its keep by being willing to say that plainly. We were also clear about what the scoring is and isn't: it's a disciplined way to compare options using the best information available now, not a guarantee of outcomes. As the company learns from its first pilots, the scores — and the priorities — should be revisited.

What the company received

The deliverable was a prioritized shortlist of use-cases with the scoring visible, a roadmap that sequenced two low-risk pilots first, and a clear rationale for what to defer and why. Instead of a scattershot "AI initiative," the team left with one page they could take to the board: here's where we'll start, here's what we're deliberately waiting on, and here's how we'll know if it's working.