Where should a business start with AI – 5 steps to choose your first problem

The most common question is not “which AI tool should we use” but “what should we start with”. Choosing the right first task matters more than choosing the right tool.

Step 1. List the repetitive work

Ask each department: what do you do over and over every day or every week? Note who does it, how long it takes, and what goes in and comes out. No need to think about AI at this step.

Step 2. Score on two axes

For each task, score the benefit (hours saved, errors reduced) and the difficulty (is the data available, do several systems need connecting). Prioritise high benefit and low difficulty.

Step 3. Check the data

AI needs documents to work from. If the price list lives in one person's head, or the process has never been written down, write it down first. This is usually the most labour-intensive part.

Step 4. Decide how to measure

Pick a number before you start: customer response time, tickets handled per day, hours spent on reports. Without a starting number you can't tell whether AI helped at all.

Step 5. Run a small pilot

Run it with one team, one channel or one type of document for a few weeks. Expand only once you see results. If it doesn't work, you stop at low cost.

Signs of a good first problem

• Happens often, every day or every week.
• Follows fairly clear rules rather than gut feeling.
• A single mistake has no serious consequence, and someone can check the result.
• The documents you need already exist as text.

You can use the 12-question self-assessment to see how ready your business is.