Should Agencies Disclose AI Use to Clients? Explain the Work and the Value

Agencies / AI and human judgment / Client trust

Yes, explain the AI.
Explain the people, too.

Where human imagination and AI work well together, the value can compound. Clients deserve to understand how that collaboration serves them and who remains responsible for the result.

By Rob Urban · PaperBoat Media

The direct answer

Agencies should explain material AI use when it affects the agreed process, client information, deliverables, rights, review or commercial terms. The conversation should make the human responsibility and value clearer. A client deserves to understand what they are buying.

What the client deserves to understand

The client is allowed to ask what the agency is doing.

If I hire an agency, I want it to use good tools. I also want to know who understands the assignment, who is checking the work and who will tell me when the attractive answer is wrong. Those expectations fit comfortably together.

AI can expand the amount of exploration a team can do and reduce time spent on tasks that do not require its best judgment. I see considerable value in that. The interesting question is what the team does with the capacity it gains. Does it investigate more carefully, consider better alternatives and spend more time understanding the client, or does it simply produce more things for the client to approve?

That is the conversation I would want an agency to be able to have. If using AI makes the value proposition difficult to explain, the value proposition may have been relying too heavily on the apparent difficulty of production. The agency has an opportunity to show the thinking that production was supposed to serve.

The disclosure problem

Three conversations are often mistaken for one.

Process, permission and public meaning require separate decisions
ConversationWhat it should establishExample
Client process disclosureHow AI is used within the agreed service and where people review the work.The agency uses an approved tool to develop concept variations, then selects and develops the final direction.
Permission and handlingWhether particular information, assets or identities may enter a system and under what conditions.A confidential launch brief is kept out of an unapproved tool.
Public-facing disclosureWhat the audience needs to understand about the final asset, subject to applicable requirements.A synthetic person should not be presented as a real customer offering genuine experience.

A general sentence saying “we use AI” does not settle any of these questions in sufficient detail. Nor does a client’s approval of one use authorize every later use. The important information should be attached to the actual work.

A decision made visible

Show one decision that the tool could not responsibly make for the client.

Consider an illustrative campaign development process. AI helps the team produce several directions, organize source material and explore alternative explanations. The agency then discovers that the most appealing direction implies a product capability the client does not offer. A human reviewer rejects it and changes the brief.

The valuable contribution is not the number of versions. It is noticing the gap, understanding its commercial consequence and choosing a defensible direction. A client should be able to see that decision in a review: what was considered, what was rejected, why the chosen approach fits and what still needs validation.

This also gives credit to the people whose work can disappear behind a polished deliverable. The strategist who asks the awkward question, the subject specialist who catches the error and the account lead who resolves a conflicting expectation are contributing real value. AI can give those people more room to do that work. The agency should be able to explain how.

Before production

Explain the approach before the client has to investigate it.

The most useful place for the first conversation is usually the proposal, onboarding or scope discussion. Describe the kinds of assistance the agency expects to use, the information controls, the review responsibilities and the circumstances requiring additional approval. A client with its own policy can then identify a conflict before production begins.

The 4As’ agency policy template includes disclosure to clients and partners and asks agencies to define approved uses. That is an industry reference point, not a substitute for the agency’s own decisions or the terms of a particular engagement. Its separate content-provenance guidance also recognizes that different AI applications raise different transparency questions.

The agency should be able to answer a direct question honestly. If a client asks whether an image, voice or passage was generated, a vague appeal to “our proprietary creative process” is unlikely to make the relationship stronger. If the team does not know the provenance of a supplied asset, that uncertainty needs to be resolved or clearly identified.

An ordinary-language explanation

Make the explanation specific enough to be useful.

The wording needs to match the actual engagement. An agency should not promise a review process it does not perform, or make a broad confidentiality assurance that its tool settings and agreements do not support. The 4As resources linked below provide useful policy and provenance considerations; the agency and its advisers still need to settle the terms appropriate to its work.

The client can then ask a better question than “Did AI touch this?” They can ask how the process improves the outcome, where human review happens and what accountability remains with the agency. The answer should be intelligible without requiring the client to become an AI procurement specialist.

The fee conversation

A faster task still needs an honest commercial model.

AI can change the time required for production. Agencies should be prepared to explain how that affects scope, capacity, quality and price. A fee based on an agreed deliverable or business responsibility is a different arrangement from billing actual hours. The explanation should match the arrangement the client purchased.

For time-based work, report time according to the agreement. Do not describe time the team did not work as labor it performed. For a fixed-fee engagement, explain the scope, expertise, review and accountability included in the price. Efficiency can support better economics for both parties, but it should not depend on a misleading account of the process.

Value-based language is not a license to avoid discussing value. Show what improved: more useful alternatives examined, a shorter review cycle, stronger consistency, a better-supported recommendation or more attention to difficult decisions. A claim that “we now work ten times faster” may be less useful to a client than a clear explanation of what the additional capacity makes possible.

From policy to practice

Put the agreement where the team can use it.

  1. Agree the use

    Record the approved purpose, tool category, information boundary and responsible owner.

  2. Check the inputs

    Confirm what may be entered or uploaded. Include client restrictions and the provenance of supplied assets.

  3. Review the output

    Use the expertise appropriate to the claim or medium. Check sources rather than accepting a confident explanation.

  4. Escalate material changes

    Ask for the required approval when the use changes the asset’s meaning, identity, confidentiality or agreed scope.

  5. Keep a proportionate record

    Retain enough information to explain important decisions and respond to a later question without turning every task into bureaucracy.

Legal requirements, contracts, platform rules and sector obligations can differ and change. The agency’s legal and compliance owners should determine which requirements apply to the actual work. This article offers a client-relationship and operating framework; it does not declare a universal disclosure rule.

The agency’s future

Will clients decide they can do it themselves?

Some will. A client may reasonably bring a task in-house when a tool makes that practical. Agencies have always had to earn work that clients could attempt themselves. The durable question is whether the agency improves the decision and result enough to justify its involvement.

A mature response is to make the division of work more useful. The client might use AI for routine internal drafting while the agency handles positioning, high-stakes creative development, research interpretation or a complicated launch. Another client may want the agency to design the workflow and train the team. Those are commercial choices worth discussing openly.

If disclosure exposes that a service contains little judgment, review or differentiation, the answer is to improve the service. If the agency contributes substantial expertise, clarity about the process can make that expertise easier to recognize. AI can help people do more of what they do best, but the agency still needs to show why that work matters.

Questions behind the agency AI conversation

Should an agency tell clients it uses AI?

My recommendation is yes for material uses within the service, with clear scope and expectations. The detail should reflect the task, client policy, agreement and applicable requirements.

Does every minor AI-assisted action need a separate message?

A well-defined policy can cover routine approved uses. Material changes, restricted information or uses requiring specific permission need the appropriate review rather than an assumption that a blanket sentence covers them.

Is telling the client the same as labeling every public asset?

No. Client process disclosure and public disclosure address different audiences and questions. Determine public-facing requirements and potential deception for the actual asset.

Should AI automatically reduce agency fees?

It can change cost and scope, but the effect depends on the pricing model and value delivered. Hours, fixed deliverables and advisory responsibility should be explained honestly under the agreed arrangement.

Can disclosure strengthen the agency relationship?

It can when it makes the process, judgment and responsibilities clearer. It will not compensate for weak work or claims the agency cannot support.

Reading behind the argument

Scroll to Top