Most businesses know they use AI. Almost none can say exactly where it touches a customer. Fewer still can say whether that involvement is visible to the person on the other end.

This exercise takes ten minutes. No software needed. It surfaces the gaps you did not know you had.

How it works

Open a blank page. Draw three columns.

Column 1: Where does AI touch a customer in your business?

List every place. Your website chatbot. The contact form. The AI tool that helps draft your client emails. The summary tool on your sales calls. The content marketing copy. The proposal generator. The recommendation engine. Everything.

One row per touchpoint. If you are unsure whether something counts, it counts. The point is to be thorough, not selective.

Column 2: Is AI involvement disclosed?

One word per row. Yes, No or Unsure.

Be honest with yourself. "It is implied" does not count. "They probably know" does not count. If a customer would be surprised to learn that AI was involved in that interaction, the answer is No.

The test is simple: would you be comfortable explaining the AI involvement to that customer face to face? If the answer is anything other than a confident yes, write No or Unsure.

Column 3: What data does that AI tool process?

One sentence per row. Customer names. Email content. Voice recordings from calls. Financial figures. Browsing behaviour. Personal details from forms. Whatever the tool ingests to do its job.

This column does two things. It shows you what you would need to explain if a customer or a regulator asked. And it surfaces the conversations you need to have with your AI tool vendors about where that data goes, who can access it and how long it is retained.

Reading the pattern

The audit is in the gaps. Look at your Nos and Unsures.

Each No is either a disclosure to add (a sentence on a webpage, a line in an email footer, a chatbot greeting that says "You are speaking with an AI assistant") or a decision to make about whether to keep using that tool without disclosure.

Each Unsure is a conversation to have. With a vendor, with a colleague, with your legal adviser. Unsure means nobody in the business has asked the question yet.

Do this once. The pattern of Unsures and Nos is your action list. It works whether you are preparing for the EU AI Act, tightening your data governance or simply building trust with the people you serve.