There is a pattern we see repeatedly across mid-market companies in industrial, logistics, and regulated sectors.
Leadership has decided that AI matters — they’ve seen the coverage, they’ve heard it from their board, some have already run pilots. The question they come to us with is usually some version of: “We know we need to do something with AI. Where do we start?”
That question, while understandable, leads organisations toward the wrong kind of work.
The tool-first trap
When “where do we start?” is answered by looking at available AI tools, companies end up with a catalogue problem. The market for AI products is now enormous and growing weekly. There is a tool for almost every function. CRM enrichment, demand forecasting, document summarisation, scheduling optimisation, quality control vision systems — all of these exist and work reasonably well in the right context.
But selecting tools before you know where leverage lives in your business leads to two outcomes: underinvestment (a few productivity tools that help individuals but don’t move the business) or overinvestment (a costly enterprise platform that gets partial adoption and produces modest returns).
Neither outcome is good, and both are common.
The right question
The more useful question is: “Where in our business model does AI change the economics?”
This question forces a different kind of analysis. Not “what can AI do in general?” but “what decisions or processes, if improved by AI, would have material impact on our margin, our capacity, or our competitive position?”
This is a business model question before it is a technology question.
For a chemical distributor, the answer might be pricing — the ability to reprice hundreds of SKUs rapidly in response to input cost changes, without the three-week lag that currently gives competitors an advantage. For a logistics company, it might be dynamic routing or carrier selection under real-time constraint changes. For a bank, it might be credit decisioning speed in SME lending, where time-to-decision is a differentiator.
The point is that the answer is specific to you. It depends on your cost structure, your competitive constraints, your operational bottlenecks, and where decisions are made slowly.
How to find the leverage point
We typically use a three-step frame with clients:
1. Map where decisions are slow, error-prone, or heavily manual. These are the places where AI typically changes economics most significantly — not because AI is smart, but because speed, accuracy, and scale compound over time in a way human labour cannot match for repetitive analytical tasks.
2. Estimate the business impact of a 30–50% improvement in each area. Not a precise ROI model — just a directional assessment of what it’s worth to move faster, reduce error rates, or handle more volume with the same team.
3. Assess feasibility and data readiness. Some leverage points are real but out of reach because the data quality, systems integration, or organisational capability isn’t there yet. Others are closer than they appear. This triage step is essential before committing to anything.
The output of this process isn’t a tool selection. It’s a prioritised view of where AI investment will produce the highest-confidence return — and a sequenced path for getting there. For a worked example of what this produces, see the AI Executive Assistant — a leverage point that surfaces repeatedly in mid-market operations.
Why this matters now
The companies that build this view early will compound their advantage. AI investments that connect to specific business outcomes create operational capabilities that are hard to replicate. Competitors who move later will be playing catch-up against an organisation that has not only implemented AI, but has integrated it into how they actually operate.
The first-mover advantage isn’t just speed to deploy. It’s the accumulation of proprietary data, refined models, and operational muscle that compounds over time.
That’s why the question to answer now isn’t which tools to buy. It’s where your leverage is.