Tell clients how you use AI before they ask

Why clients are asking, and what a bad answer costs

Clients have moved past the vague question of whether you use AI. They are now asking how you control it, and many firms cannot answer that cleanly. When a client pops the question in a meeting, the pause can be audible. clients have stopped asking whether you use AI and started asking how you control it

A written answer prevents the improvisation that erodes trust. One consultant describes a client asking, “Are you running our stuff through AI? And if you are, where does it go?” The owner on the other side of that call froze, not because he was hiding anything, but because he had never written the answer down. He could hear himself improvising in real time. You tell them the truth, in writing

The same source gives the direct answer: tell them the truth, in writing. A consistent written disclosure is easier for your team to share, less likely to vary from one conversation to the next, and easier to keep accurate as your tools change.

There is also a market expectation at work. In a survey of US buyers, 84% said they want AI content labeled. Yet only 20% of firms always say when they use AI, and a third never do. survey of US buyers That gap is a reputational risk. The same source argues that telling people is not the risk. Being found out is.

A bad answer costs more than a good one. Over-disclosing can invite the client to ask why your fee has not dropped, because they read it as a sign the work has become cheaper. Under-disclosing reads as concealment when the client learns about your AI use from a competitor or a tell in a deliverable, and that is harder to recover from than the original conversation would have been. over-disclosing and under-disclosing The goal is a calm, factual middle ground.

What to include in a one-paragraph disclosure

Your disclosure should answer three practical questions:

• What do you use AI for?
• What do you never put into AI?
• Who reviews and is responsible for AI output?

Keep the focus on data handling, not on tool names. Tool names change, but the commitments you make about client data should be stable. A paragraph built around data handling also avoids the trap of over-promising about a specific vendor.

A workable paragraph might read: “We may use AI tools to help draft routine documents, summarize internal research, or improve internal workflows. We do not enter client confidential information into any AI tool without client consent. Any AI-generated output is reviewed by a person before it is used in client work. If you have questions about our use of AI, contact [name].”

That paragraph is clear, short, and does not promise more than you can deliver. Do not claim you “never use AI” if you use it for spellcheck, autocomplete, or meeting transcription. A regulator may check whether your stated practice matches your actual practice. regulators treat AI tools the same as any other tool If you say you review everything, you must actually review everything. If you say you do not share client data, you must make sure no one on your team is pasting a contract into a free tool.

How to handle the follow-up: 'Which AI tool do you use?'

When a client asks for a specific product name, they are usually asking about risk, not curiosity. They want to know whether their data is being used to train someone else's model or stored somewhere they did not agree to.

Answer with the category and the safeguards first. For example: “We use a small set of tools for drafting and research. We only use tools configured so that client data is not used to train the underlying models. A list and the data handling details are available on request.”

This approach works because it answers the actual concern without locking you into a product name that may change next quarter. If you do name a tool, be prepared to describe how it handles data, including whether your team can turn off training, where data is stored, and how long it is retained. If you do not know, say you will confirm in writing and then do it. That is far better than guessing.

A written disclosure lets your team give the same answer whether they are asked in a meeting, over email, or at a conference. The alternative is a different answer from each person, which looks like inconsistency rather than confidence.

When to get client consent instead of just notice

Notice is enough for general AI use that does not involve client confidential data. But if you plan to upload contracts, health records, financial statements, or other regulated or privileged information, you may need consent, not just disclosure.

Check your client contracts and any professional guidance before assuming you are exempt. Check your client contracts and professional guidance Some engagements require written consent before you use a third-party service on client data. Others may prohibit it altogether, or require you to use tools that have been assessed for security and confidentiality.

Also consider your industry rules. Regulators treat AI tools the same as any other tool, so a claim about AI is a claim about your practice. regulators If you tell a client you never share their data, but you use a generative AI tool that trains on prompts, that could be a problem.

Jecovia works alongside your attorney to determine what your professional rules and client agreements require. When in doubt, ask before you paste.

Where to publish it: engagement letters, portals, your website

Put the disclosure where clients already look. Add a short paragraph to your engagement letter or terms of service. That creates a record without extra meetings, and it shows the client you have thought about the issue.

Post the same paragraph on your client portal and your website. If a client asks years later, you can point to the version they agreed to. If you have a privacy notice, link to the AI disclosure from there.

Give your internal team the same paragraph for emails and calls. A client should not hear a different answer from your office manager than from you. Consider adding it to your internal knowledge base or staff handbook.

If you publish a longer AI policy, link to the one-paragraph summary so clients can read the details if they want. The goal is not to hide the policy, but to make the important part easy to find.

Keep the disclosure current as your tool list changes

Your AI tool list will change. Your disclosure should change with it.

Review the paragraph quarterly or whenever you add or remove a tool. If a new tool uses client data differently, update the disclosure before you deploy it, not after a client asks.

Assign one person to own the disclosure. That person should know what tools are in use and what data each one touches. They should also know who to ask if a new feature changes data handling.

If you need help evaluating a new tool before you write it into your disclosure, see Jecovia's AI & Emerging Technology services.