Branding in the Age of AI: What Should Clients Pay For Now?

A business can commission fifty brand directions before its leadership team has agreed what the business should stand for.

That is the tension at the heart of branding in the age of AI. Producing possibilities is becoming easier. Choosing a direction, proving its relevance and carrying it into everyday experience still require work.

For clients, the opportunity is better use of their budget. For agencies, the challenge is explaining precisely where their contribution improves the result.

Our argument is simple: human involvement should earn its place through the quality of the decisions it enables.

AI is changing more than the production process

The World Federation of Advertisers’ September 2026 survey found that AI use extended beyond content into research, insights, strategy and planning. Its findings represent 54 senior respondents from 46 major brand owners, rather than the whole business population. Nevertheless, they show that sophisticated buyers are already applying AI to work agencies have traditionally sold as expertise.^[1]

Agencies should take that seriously. Summarising research, drafting messages, exploring visual territories and adapting approved material are all candidates for AI assistance. The appropriate level of automation depends on the task, available evidence and review process.

Clients can reasonably ask why a proposal still allocates extensive production time to work now completed more efficiently. They can also ask what happens to the time saved: does it reduce the fee, improve research, support testing or simply disappear into the agency’s margin?

Those questions belong in the commercial conversation.

Human involvement has to justify itself

The premise that clients always need more human branding deserves scrutiny.

A field experiment published in Organization Science in June 2026 studied 791 professionals at Procter & Gamble working on product innovation challenges. Individuals using AI matched the performance of teams working without it. AI also helped participants integrate commercial and technical perspectives.^[2]

This was a study of particular innovation tasks within one company. It does not establish that an AI-assisted founder can replace an entire branding team. It does, however, challenge the assumption that adding people necessarily adds value.

An experienced internal team adapting an established campaign may need limited external involvement. A founder testing an early proposition may benefit from AI-supported exploration followed by a focused expert review.

More intensive human work becomes worthwhile when there are unresolved choices: conflicting stakeholder priorities, unfamiliar markets, weak customer evidence or promises the organisation may struggle to deliver.

The useful question is: which decisions require expertise, direct observation or accountable approval?

Evidence matters more when answers sound convincing

AI can help organise interview transcripts and identify possible themes. It cannot turn invented interviews into customer evidence.

NIST’s Generative AI Profile identifies confabulation: systems can confidently produce false information, including misleading explanations and citations.^[3] For branding, the practical implication is straightforward. A convincing description of an audience remains a hypothesis until supported by evidence about that audience.

A separate experiment involving BCG consultants found that AI improved performance on tasks within its capabilities but reduced correctness on a task outside them. That experiment used GPT-4 in 2023; it should inform testing practices, rather than serve as a performance forecast for today’s tools.^[4]

Human judgement is valuable here when it is informed and testable. A strategist’s intuition can also be wrong. Interviews, observation, commercial data and documented reasoning give both people and AI something more reliable to work with.

Distinctiveness needs an independent starting point

In a 2024 Science Advances experiment, access to AI-generated ideas improved evaluations of individual short stories, while the AI-assisted stories became more similar to one another.^[5]

The study concerned short fiction, not brand identities. The implication for branding is a hypothesis worth testing: teams drawing on similar generated references may produce individually competent work that becomes difficult to distinguish collectively.

Agencies cannot answer that risk simply by claiming to be creative. They need to demonstrate where a direction comes from: a product behaviour, customer frustration, organisational conviction or cultural observation that competitors cannot adopt without contradiction.

Cultural understanding also requires specificity. “An African audience” is an inadequate brief. Language, location, income, context and expectations can change what a message means. Use people with relevant knowledge to investigate those differences, and test important interpretations with intended audiences. Neither a locally based person nor a model deserves automatic authority over everyone’s experience.

 

What clients should pay for

A stronger proposal makes four contributions visible.

Discovery that resolves a real question. What must be understood before the business chooses a position? Specify the research, its limitations and the decision it will inform.

A strategic recommendation with reasons. Which audience, promise and competitive position should the organisation prioritise? Show the alternatives considered, evidence used and trade-offs accepted.

A system people can implement. An identity needs rules, examples, usable files, messaging and ownership. Its value includes whether internal teams and partners can apply it consistently without repeatedly returning to the designer.

Validation and improvement. Agree how to assess the work: comprehension, recognition, usability, adoption or relevant commercial indicators. Distinguish what the agency can influence from outcomes affected by pricing, distribution, service quality and other factors.

Production remains legitimate paid work. Its price should reflect scope, craft, complexity and responsibility. Clients should request clarity about AI use, third-party costs, review effort and how efficiency gains affect the engagement. Agencies should avoid dressing routine generation up as original research or bespoke strategy.

 

Put the collaboration model in the brief

Clients contribute business priorities, operational truth, customer access and decisions about acceptable risk. Agencies contribute research design, interpretation, creative direction, implementation expertise and a reasoned recommendation. AI can support defined tasks across that process.

These responsibilities should be explicit before work begins.

Agree which tools may receive which information, where AI assistance will be disclosed and who verifies factual claims. Name the people responsible for final approval. Identify any specialist review needed for rights, privacy, technical accuracy or sensitive communications.

This transparency addresses a documented concern. In WFA’s 2026 survey, 69% required agencies to disclose AI use, while only 13% reported substantial visibility into it.^[1]

Ownership also needs attention. The US Copyright Office’s January 2025 report distinguished sufficient human authorship from merely providing prompts, while recognising that AI assistance does not automatically prevent copyright protection.^[6] That is a US-specific conclusion; rights decisions need review for the relevant jurisdiction and intended use.

Apply human effort where the consequences justify it

Consider three illustrative situations, rather than claimed client results.

A startup developing its proposition: AI helps generate alternatives and challenge assumptions. The founder and strategist investigate customer needs, test the strongest options and choose a promise the business can fulfil. The deliverable is an evidenced direction with a practical first phase.

An organisation improving a workplace: AI supports early visual and messaging exploration. Designers and relevant specialists inspect the site, evaluate legibility and accessibility, and test whether people understand the signs. A compelling mock-up is only an early stage of a usable environment.

 

 

A marketing team scaling communication: AI drafts variations from approved messages and source material. Editors check accuracy, tone and cultural fit. The team tests performance and records what it learns. Review effort increases where a mistake would have greater consequences.

At Argillic, these decisions connect Strategic Architecture—what an organisation stands for—with Experience Systems—how people encounter it—and Communication Systems—how its meaning travels. AI can assist across all three. The work is making sure they reinforce one another.

Ask for a better decision

Before commissioning branding, ask the agency: what will we understand, decide or implement better because you are involved?

Before deploying AI, ask the same question of the tool.

The answer should shape the team, fee and process. Some tasks will need fewer people. Others will justify deeper research, specialist judgement or more careful testing. The objective is a brand the organisation can explain, deliver and sustain.

If you are considering how AI should fit into your brand strategy, communication or customer experience, speak with Argillic about the decisions your project needs to resolve.

Sources

1. World Federation of Advertisers (30 September 2026). State of AI in Marketing: public survey findings. 54 respondents from 46 major brand owners.
2. Dell’Acqua and colleagues (12 June 2026). The Cybernetic Teammate. Organization Science. Published analysis: 791 P&G professionals.
3. NIST (July 2024). Generative Artificial Intelligence Profile, AI 600-1. See section 2.2 on confabulation.
4. Dell’Acqua and colleagues (11 March 2026). Navigating the Jagged Technological Frontier. Organization Science. Experiment used GPT-4 in June 2023.
5. Doshi and Hauser (12 July 2024). Generative AI and individual creativity versus collective diversity. Science Advances; short-fiction experiment.
6. US Copyright Office (29 January 2025). Announcement of its report on AI output copyrightability. US-specific finding.