Should an agency tell clients it uses AI?

A client recently audited their agency's deliverables, found unreviewed AI output, and posted that they were "at a loss for words" — asking the internet whether to fire them. A hundred-plus comments said some version of yes. Read the thread closely, though, and the offence wasn't the tool. It was that nobody at the agency had read the work, and that the client found out by accident.

What clients actually punish

Clients fire agencies for two things in these stories: quality that fell below what they're paying for, and the feeling of having been deceived. AI is the occasion, not the cause. An agency that ships sharp, accurate, on-voice work and is matter-of-fact about its method almost never features in these threads.

The honest framing

"AI-assisted, human-approved" is the accurate description of a well-run modern agency, and it's a strong pitch rather than a confession: machine speed on production, senior judgement on everything that ships, your accountability on the result. Put it in the proposal and the onboarding deck. Specify what AI drafts (captions, variants, first-cut visuals), what humans do (strategy, editing, approval, reporting), and what the client sees (only approved work, with an approval trail).

White-label is presentation, not deception

Worth being precise here, because it's our own category: white-labelling means the platform your clients see carries your agency's brand — your logo, your domain. That is normal commercial practice, like a bakery selling under a café's name. It is not a licence to misrepresent your process when a client asks how work gets made. Brand the service; tell the truth about the method.

The client who demands zero AI

They exist, and they're entitled to the preference — price the fully-manual service accordingly, or decline gracefully. What you cannot do is agree and then quietly generate anyway. Every version of that story ends in the same Reddit thread.

Disclosure costs a sentence. Discovery costs the account.