What to tell clients about AI (and what not to)
In brief
The wrong AI conversation with a client creates a problem you'll spend months managing. The right one positions you as the person who gets it before everyone else does. Here's the script for both scenarios.
Contents
Every agency is having some version of this conversation: a client asks how much AI is involved in their work, and the account lead either freezes, deflects, or overclaims. None of those serve the relationship.
The problem is not that clients are asking — it is that agencies have not developed a clear, honest answer. Here is one.
The honest framing
AI is a production tool in our workflow, the same way design software or a research database is. It helps our team move faster on first drafts, synthesis, and adaptation. The judgment, strategy, editorial voice, and approval chain are still entirely human.
That is the full answer. It does not oversell, it does not hide, and it accurately describes what is actually happening in almost every agency using Claude responsibly.
Clients who ask "are you using AI?" are usually asking one of three underlying questions:
Am I paying for work a machine did? — The right response is to clarify what they are paying for: expertise, judgment, and accountability. Claude drafts; your team directs, revises, and is responsible for what goes out.
Is the work going to be generic and off-brand? — This is the real concern. Address it directly: "Our process includes a client context document that gives Claude your brand voice, constraints, and past feedback. We review everything before it leaves the team." That is more reassuring than any abstract answer about AI quality.
Are you going to charge me the same for half the work? — See the billing section below.
What not to say
"We're experimenting with AI." This sounds evasive and implies you are using the client as a test subject. If you are using Claude in production work, say so.
"AI writes everything and we review it." This overstates Claude's role and removes your team from the story. The work is yours. Claude is a tool.
"Our AI never makes mistakes." No. Claude hallucinates, especially on specific facts, statistics, and citations. Clients who use AI themselves know this. If you deny it, you lose credibility. The honest version: "Claude sometimes produces errors that require human review — which is why our process always includes a review step before delivery."
The hallucination conversation
If a client specifically asks about hallucination or accuracy, you need a real answer.
Claude produces plausible-sounding text. For most creative and strategic work — copy, briefs, frameworks, analysis of information you provided — accuracy errors are rare and the review step catches them. For tasks that require factual accuracy about the real world — statistics, citations, competitor specifics, current events — Claude needs to be grounded with provided source material, or the output needs human fact-checking.
The practical rule: if a fact in Claude's output matters and you cannot verify it from a source you provided, verify it externally before it leaves your team. This is not unique to AI. It is editorial hygiene.
The billing conversation
Clients sometimes push on pricing when AI is involved. The right response is to anchor on outcomes, not process.
"We charge for the strategic judgment, brand expertise, and accountability that makes the work effective — not for the hours spent in any particular tool. Our use of AI means we can deliver faster without compromising quality. The expertise behind the output is the same."
If a client pushes further, offer to walk through the actual process. Show them the brief, the angle development, the revision rounds. Most clients, once they see the work that shapes and directs the AI output, stop asking about AI pricing. The process is defensible because it is real.
What proactive disclosure looks like
Some agencies are choosing to proactively disclose AI use rather than wait to be asked. The agencies that do this well frame it as a process strength, not a confession.
"Our workflow now includes AI-assisted first-draft production, which means we can get to a reviewable draft faster and spend more of our time on strategy and revisions. Everything still goes through our senior team before it reaches you."
That framing — AI makes us faster at the mechanical work so we can do more of the high-value work — is accurate, confident, and client-friendly. It also positions the agency as operationally sophisticated rather than reactive.
Further reading
- How enterprises are driving AI transformation with Claude — what to tell clients based on real patterns