A few years ago, I needed a visual representation of a value chain. It was a diagram showing how one activity led to another, not a campaign or a new identity. I sent the request to our internal design department and discovered that they could not produce it. Whether the obstacle was capability, process or a narrow view of what counted as design no longer matters. Someone eventually assembled the value chain in PowerPoint, and we used that chart. The incident seemed like a minor organisational absurdity: the company had designers, software and brand guidelines, yet a useful picture could not pass through the system.

A generative system could now complete the same request before the design department acknowledged the email. It could propose visual approaches, render a polished diagram and adapt it for several channels. Removing such delays is a real gain. The awkward part is what follows. If the system can also draft the campaign, create the images and populate the social calendar, what work remains for creative marketing?

The usual answer assigns strategy to people and execution to machines. That division assumes a clarity many organisations do not possess. Marketing strategies often amount to instructions to launch a campaign or refresh a brand, without establishing what the resulting work should make customers understand. Creative expertise has therefore been associated with the scarce ability to turn an incomplete instruction into an artefact. Production required time and money, which limited how many possibilities a company could pursue and forced someone to choose among them. The PowerPoint chart showed that difficulty of execution was not a reliable measure of an idea’s usefulness. Even so, the cost of making creative work prevented every plausible idea from becoming a campaign.

Research on design gives this change a useful structure. Hou and colleagues divide the creative process into an ideation stage, where designers generate possibilities, and an implementation stage, where they select and refine a solution. In two experiments involving graphic-design tasks, generative AI improved the creativity of ideas produced by both novice and expert designers. Its effects during implementation depended on expertise. Novices continued to benefit, while experts spent more time revising and reworking AI contributions that disrupted their established methods.¹ The study concerns particular design tasks and cannot settle the future of marketing. It does show that creative work contains activities that AI affects differently.

For marketing, the decisive pressure appears once ideation becomes abundant. Twenty headlines or visual routes can be produced before a team has established what would make any of them appropriate. Another option adds little when the existing options are already competent. The scarce ability is discrimination: recognising why one possibility belongs to this organisation, at this moment, and why the others should be discarded. Such judgement depends on information that may be absent from a prompt, including what customers have experienced and which promises the company has failed to keep. It also requires the authority to reject an attractive idea when publishing it would create a claim that the organisation cannot support.

A marketer’s personal preference is not enough to justify choosing one creative option over another. The decision has to be tested against what the communication asks its audience to believe and what the organisation can show in return. Before approving a campaign about effortless service, marketers should compare the promise with the points where customers still encounter delay. Generative AI can produce alternative expressions and help test how clearly each one communicates the claim. It cannot remove the discrepancy between the promise and the service.

Zhou and Lee studied more than four million artworks made by over 50,000 users of a text-to-image platform. After adoption, users produced 25 per cent more work, and the likelihood of receiving a favourite per view rose by 50 per cent. Peak content novelty increased, but average novelty declined.² The finding complicates a simple story of creative displacement. Generative tools can help individuals make more highly valued work while the average output moves towards greater similarity. In a marketing department, the danger is therefore easy to miss. Each asset can be polished and technically distinct even when the choices behind the assets are becoming more conventional.

A brand makes the problem institutional. A system supplied with tone-of-voice rules, approved claims, product information and earlier campaigns can speak in a recognisable style at a scale no team could match.Hancock, Naaman and Levy use the term AI-mediated communication for messages that an intelligent agent helps produce on a communicator’s behalf.³ Their framework makes the problem one of agency: the system participates in shaping how the sender appears to others.³ Applied to a company, factual accuracy is only part of the issue. Fluent brand language should also record an act of organisational attention.

Customers interpret company speech partly through the assumption that someone chose to speak. A product announcement suggests that the organisation considers the change worth noticing. A promise of simplicity invites customers to compare the claim with the service they receive. The statement has weight because it exposes the company to contradiction. Automatic production can weaken that connection without producing a single false sentence. A brand can generate a claim for each audience, revise it quickly and continue publishing when nothing inside the organisation has changed. The calendar remains full while each communication provides less evidence that the company noticed anything new.

Creative marketing becomes more consequential under these conditions, although its visible craft may occupy less of the process. Its work begins in sales conversations, complaints, product limitations and disagreements that are not already contained in the brand archive. Marketers must decide which of those encounters alters what the company can credibly say. AI can help them test an idea, expose a weak assumption or make a useful diagram without waiting for scarce production capacity. The tool creates material on which judgement can operate. It cannot decide whether a claim arose from evidence or whether the organisation accepts the obligation created by publishing it.

My PowerPoint chart deserved to exist because it helped people understand the value chain. A better tool could have made it faster and clearer. The mistake would be to treat the improved picture as evidence that the company had understood the process depicted in it. The same mistake now threatens brand communication at scale. When expression was expensive, organisations sometimes confused the difficulty of making an asset with the value of the idea behind it. Cheap execution removes that confusion, but it introduces another: the ease of saying something can look like a reason to say it.

Creative marketers will not justify their role by preserving delays or claiming a monopoly on imagination. Their harder task is to maintain the connection between a company’s language and its conduct. When a brand can say almost anything, creative judgement determines what the organisation has earned the right to say.

Notes

  1. Jinghui (Jove) Hou, Lei Wang, Gang Wang, Harry Jiannan Wang, and Shuai Yang, “The Double-Edged Roles of Generative AI in the Creative Process: Experiments on Design Work,” Information Systems Research, published online October 3, 2025.
  2. Eric Zhou and Dokyun Lee, “Generative Artificial Intelligence, Human Creativity, and Art,” PNAS Nexus 3, no. 3 (March 2024): pgae052.
  3. Jeffrey T. Hancock, Mor Naaman, and Karen Levy, “AI-Mediated Communication: Definition, Research Agenda, and Ethical Considerations,” Journal of Computer-Mediated Communication 25, no. 1 (January 2020): 89–100.