A client once forwarded three of my follow-up emails back to me and asked, very politely, whether I had written them. I had not written them. I had written the prompt. Nothing had gone wrong, the work was fine, better than fine. But I spent the next forty minutes on that call explaining something I could have covered in four sentences up front.
That is the whole case for the client-facing AI disclosure email. It is not a confession. It is a boundary document, and it goes out before the workflow, not after somebody notices.
Why the Email Has to Land Before the Workflow Does
The question I hear most from founders is some version of should I tell clients I use AI. The answer is yes, and the more useful question is when. Before you flip the switch, or after.
Before sets the terms. You get to define the edges: which tools touch their work, which parts stay human, what happens to their data. After lets the client set the terms, and clients set worse terms when they feel like they caught something. An AI disclosure email to clients is not a legal document. It is the difference between a client who feels informed and a client who feels handled.
There is a second reason that has nothing to do with the client. Writing the email forces you to actually know what is running in your business. When I sat down to list every tool that touched client work, I found a call recorder still storing transcripts in a workspace that had outlived the project it was set up for. Nobody had asked me about it. It was still a problem.
So make the list first. Call recording: Fireflies, Fathom, Otter. Drafting: Claude, ChatGPT. Routing and follow-up: Zapier, Make, HubSpot workflows. Plus transcription, scheduling, invoicing. If you cannot name the tools, you cannot disclose them, and you are not ready to automate the workflow yet.
The Four Things the Email Has To Say, and One It Shouldn't
There are four pieces of information a client actually needs, and none of them are technical.
First, the names. Not "AI solutions." Fireflies and Claude and Zapier. Named tools are checkable. Unnamed tools are a suspicion.
Second, what each tool touches. "It records our calls" and "it drafts internal recaps I rewrite before you see them" are completely different levels of exposure, and the client deserves to know which one they are getting.
Third, what stays human. Be specific about the deliverable and the judgment calls. This is the part that carries the relationship. Client trust in AI automation comes from being able to see the fence around it, not from a promise that the fence is strong.
Fourth, data handling. Whether their work trains anything, how long recordings live, who inside your business can see them. Two sentences is enough.
The one thing the email should not do is hedge. "We use AI to improve efficiency" is the sentence that makes people nervous, because there are no edges in it. Vagueness reads worse than the automation itself.
The Template
Solo consultants ask me for an AI disclosure template constantly, and they always expect it to be longer than it is. Here is the version I send. Plain, short, no legal language, no apologies.
Subject: How we use AI on your account
Hi [Name],
A short note on something I should have said out loud earlier. [Business] uses a small set of AI tools in the background of your work. Here is exactly what that means.
What we use:
- [Fireflies] records our calls so I am not typing while you talk. Recordings live in [workspace] and get deleted after [90] days.
- [Claude] drafts the first pass of recaps and internal notes. I rewrite them before anything reaches you.
- [Zapier] moves files between [tools] and triggers reminders. It does not touch the content.
What stays human: [strategy, the deliverable itself, the final review]. None of that gets handed to a model.
What does not happen: your work is not used to train anything, nothing goes into a public model, and [nothing leaves our workspace].
If you would rather we not run a specific tool on your account, reply and tell me which one. I will turn it off for you, no explanation needed. It is a one-line change on our end.
Happy to walk through any of this on our next call.
[Name]
That last paragraph is doing most of the work. It turns the email from a notification into a choice, and a client who has a choice does not need to go looking for one.
When a Client Pushes Back
Some will. It is usually not about the technology.
Do not argue and do not over-explain. Three responses cover almost every version of this conversation.
The carve-out: ask which specific deliverable they want done without AI, and agree to it in writing. Most clients do not want zero AI. They want one thing that is definitely yours.
The walkthrough: offer to screen share the tool and show exactly where their files go and when they get deleted. Fireflies has a settings page you can show in ninety seconds. That is usually the end of it.
The honest limit: if they need a fully AI-free backend and yours runs on automation, say so in week one and let them decide. That is a fit problem, not a trust problem, and pretending otherwise costs you both.
One thing I have noticed. The clients who ask sharp questions about your AI use are usually the good ones. The ones who go quiet and never mention it again are the ones to watch. If you work with enterprise accounts, they will hand you their own AI use policy questionnaire. Answer it in writing, keep a copy in the account file, and reuse the answers.
Turn the Email Into a Policy So You Stop Rewriting It
One email per client is a habit. A policy is infrastructure, and you want the second one. An AI use policy for a service business can be a single page, as long as it exists somewhere other than your head.
Three steps. Write a one-page inventory in Notion or Google Docs that separates client-facing tools from internal ones, and update it when you add something. Drop a two-paragraph AI section into your master services agreement or your standard statement of work, right next to the confidentiality clause, so it is part of the deal and not a surprise later. Add a line to your onboarding checklist in Airtable or Notion that says "send AI disclosure" next to "send welcome packet."
That is it. Twenty minutes of work you do once, and every new client gets the same clear answer without you writing it from scratch at 11pm. We build these documents live in the workshop at https://workshop.mastermindshq.business, and if you want to see how this piece fits into a larger operating system, the comparison page at https://www.mastermindshq.business/mastermindvsl is a good place to start.
It also sells. Clients are putting AI questions in RFPs now. Being the studio that has a written answer ready is a small, real advantage, and it has nothing to do with whether the work is any good.
If this resonates and you want to build this kind of infrastructure in your own business, the Mastermind is where we do the work live. You can learn more at mastermindshq.business.