Client work and AI: what to disclose and when
What to say, when to say it, and why it builds trust.
If you're doing client work, AI is now part of how a lot of that work gets done. Drafts written with AI, research compressed with AI, meeting summaries cleaned up with AI. Sometimes more. Which raises a question most agencies, consultancies, and freelancers haven't worked out yet: do you tell the client?
There is a working framework for that. Disclosure is sometimes required, sometimes expected, sometimes optional, and sometimes just weird to bring up.
The principle underneath
Clients are paying for outcomes and judgement. They are not paying for every keystroke to be human. You don't disclose spell-check, a calculator, or a search engine.
AI sits on the same spectrum, with one difference that matters: public perception is still settling, and the line between tool and substitute isn't widely agreed yet. So the real question isn't only "should I disclose". It's "what would this client think if they found out later?" If the honest answer is that they'd be fine with it, you probably don't need to. If the honest answer is that they'd feel misled, you do.
One case sits outside that test, because it isn't about disclosure at all. If a task that normally takes eight hours took two, billing for eight is fraud whether or not you mention AI. Bill the actual time, bill on value, or pass the saving on.
Always disclose
The contract or brief specifies human-only work
AI output presented as your own research or analysis
AI did the strategy, the concept or the core analysis
Client data went into an AI tool
The work is public-facing under a human's name
No need to disclose
Used as a search engine, with nothing in the deliverable
A draft you then rewrote substantially
Cleaning up your own meeting notes for your own use
Formatting, restructuring, fixing typos
Brainstorming where the ideas you chose were yours
If you'd be uncomfortable with the client finding out, you already know the answer.
The grey middle
Between the obvious cases sits most real work, and a few things push it towards disclosure. If the client has their own sensitivity about AI, whether policy, internal politics or competitive concern, one short conversation now prevents a difficult one later. If the output is public-facing and carries a human's name, an op-ed, an article, a talk, they should know how it was produced. And if their data went into any tool, they have a right to know which tool and what protections were in place.
The rest is genuinely a judgement call, and the useful default is to describe your standard practice once, in your engagement materials, rather than narrating every instance. Something like: "We may use AI tools to support drafting, research and analysis. All work is reviewed and finalised by us. Client data is handled in accordance with our data policy."
Have the conversation first
The simplest way to avoid the question later is to raise it during onboarding. "We use AI tools as part of how we work, mainly for drafting, research and clean-up. All output is reviewed and finalised by us. Any client information stays on tools with appropriate data protections. If you'd prefer we didn't use AI on parts of this engagement, tell us and we'll adjust."
Most clients won't push back, some will appreciate the openness, and a few will have strong views either way, all of which is better known upfront. And if one asks directly, answer plainly. "Yes, for that part. The rest was done by hand." Hedging is where trust actually breaks.
Three things to do this week
01
Have the conversation before the work, not after
Say how you use AI once, at the start, and put it in your engagement materials. It's welcomed far more often than resisted.
02
Disclose when AI did the substantive thinking
Tidying your own notes isn't the same as it writing the deliverable. The second one they should know about.
03
Don't over-disclose the boring parts
Nobody needs telling you used AI to clean up your own meeting notes. Over-disclosing makes the real disclosures mean less.
The upfront conversation
Say this once, during onboarding, and most disclosure questions never come up. Quick note on our process: we use AI tools as part of how we work, mainly for drafting, research, and clean-up. All output is reviewed and finalised by us. Any client information we use stays on tools with appropriate data protections. If you'd prefer we don't use AI on parts of this engagement, just let us know and we'll adjust. For your engagement materials: We may use AI tools to support drafting, research, and analysis. All work is reviewed and finalised by [name/role]. Client data is handled in accordance with [policy].
Conclusion
01
Always disclose when the contract requires it, or when AI did the substantive work.
02
If you'd be uncomfortable with the client finding out, that discomfort is the answer.
03
One short conversation upfront handles most disclosure questions before they come up.
Go to next