Most guides to AI implementation assume you have a data team. Most companies do not. This is for the other group: the founder, the operations director, or the department head who knows AI should be on the agenda but is not sure where to start.
The most common mistake
Starting with the technology rather than the problem. This leads to pilot projects that are technically impressive and operationally useless. The model works. Nobody uses it. The project quietly dies.
The right starting point is a specific operational problem. Not “how can we use AI” but “which task costs us the most time and follows the most consistent rules?”
Three stages
Stage 1: Find the seam. Spend two weeks mapping your highest-volume, most time-consuming workflows. For each one, ask: what are the inputs? What decisions get made? What are the outputs? Where do errors happen? This is not a technical exercise. It is an operational one.
Stage 2: Build the smallest useful version. Resist the urge to automate everything. Pick one workflow, one team, one system. Build the smallest version that actually saves time. Get it into real use. Learn from it. Then expand.
Stage 3: Build governance in from day one. Before you build anything, decide: who owns this system? How do we know if it is working? What happens when it gets something wrong? These questions are boring. They are also the reason most AI projects survive past the pilot.
What you actually need
You do not need a data scientist. You do not need a large budget. You need a clear description of the workflow you want to improve, access to the relevant data, and someone who can test it honestly.
A readiness audit takes 2–3 weeks and gives you a prioritised roadmap. A first workflow automation takes 4–8 weeks. Most clients see measurable time savings within 60 days of going live. Start here.