A client asked me last month a simple question: "Do you put my stuff into AI?" I gave him a vague answer. He could tell. He stopped the call and asked for something in writing. That was fair.
If you run a small business, an agency, or a consulting practice, you will get this question. Sometimes it comes in a sales call. Sometimes it comes in a security questionnaire, or from a client's own operations lead who has been told to check. Either way, you need a clear answer that is true and that you can back up.
Here is how I'd approach it.
Start with the honest list.
Before you write anything for a client, write down what you actually use. Most people undercount. You might use an AI writing assistant for drafts, a transcription tool for calls, a scheduling tool that suggests replies, and a browser extension that summarizes emails. Each one touches data in some way. If you can't list them, you can't describe them.
Go through your last month of work. Open your app list. Check your browser extensions. Ask your team what they've been using. Include the free tools people signed up for on their own. Those count too.
For each tool, answer four questions.
- What client data goes into it? Names, emails, financial figures, internal documents, call recordings?
- Where is that data processed and stored? Check the vendor's documentation. If you can't find it, say so and treat that as a gap to close.
- Does the vendor use your inputs to train its models? Read the terms for your plan. Free and paid tiers often differ. Write down what the terms say and the date you checked them.
- Who on your side can access it, and how long is it kept? Include deletion options if the tool offers them.
If you don't know an answer, don't guess. Write "I'm confirming this with the vendor" and then follow up. Clients respect a specific "not yet" more than a confident wrong answer.
Set rules for what stays out.
This is where clients usually relax. Tell them what you don't do. For example: you don't paste passwords or login credentials into any AI tool, you don't upload full client databases, and you don't feed confidential contracts into a tool whose terms you haven't reviewed. Make the rules short and specific. A rule like "we use AI responsibly" tells the client nothing.
Say who checks the output.
Clients also worry about accuracy and judgment, not just data. Explain that AI drafts are reviewed by a person before anything goes out, and name the role responsible. If AI helps with analysis, say how the results are verified. This is part of how you handle their data, because a wrong number sent under your name is a problem too.
Write a one-page answer.
Put it all on one page. Include:
- The tools you use and what each one handles
- Where data is stored and whether it trains vendor models
- Your rules for what never goes in
- Who reviews AI output
- How and when you'll tell the client about changes
Update the date at the top. Send it when asked, and offer to send it again when something changes. If you change a tool, update the page the same week.
Keep a running record.
The one-page answer is only as good as the record behind it. Keep a log of every AI tool and automation you use, with the data it touches and the date you last checked its terms. TechEd Analyst's web tool is built for this kind of record, so you aren't rebuilding the list from memory each time a client asks. A spreadsheet works too. The point is that the record exists and is current.
Be careful with what you promise.
Don't tell a client that your setup makes them compliant with any law or standard. That is a claim about legal outcomes, and it isn't yours to make. What you can say is that you keep a documented record, you review vendor terms, and you have a process for answering questions like this one. If a client needs a formal legal or regulatory opinion, send them to their own counsel or auditor. That's the right answer, and it's the one that protects both of you.
The short version.
When a client asks how you use AI with their data, answer with a specific list, a clear set of limits, a named reviewer, and a page you update. Be plain about what you know and what you're still checking.
That answer builds more trust than any polished sales line. Clients don't expect perfection. They expect you to know your own tools and to tell them the