Three questions to ask before you spend on AI

Most AI spending starts with a tool: a licence, a pilot, a chatbot. The money goes out, the demo looks good, and three months later nobody can say what changed. The problem is rarely the tool. It is that three questions were never asked.

1Which number should move?

Not “we want to use AI”, but one business number: sales in a region, customers kept, staff retained, hours spent on reporting. If you cannot name the number, you cannot tell whether the money worked.

2Where does the data that explains it live today?

Every number is driven by something you already record somewhere: the CRM, the ERP, timesheets, support tickets, a spreadsheet someone updates every Friday. AI can only explain the number if it can see that data. Make a list of where it lives and who owns it. That list is your real starting point.

3Who will act on what the AI says?

An insight nobody acts on is a cost. Before you build anything, decide which person or team receives the signal, where it lands (their inbox, the CRM, a Monday report) and what they are expected to do with it. If no one owns the action, the number will not move.

What good looks like

When all three are answered, the project almost designs itself: connect the data, let the AI find the pattern, put the signal in front of the person who acts, and measure the number before and after. That is the whole job, and it is the part most AI pilots skip.