Where AI belongs in the creative process, and where it doesn't

AI belongs in the creative process after the idea and before the final call. It widens the search and scales the output, while people decide what the work should say and whether it is good enough. At IW Agency, we think the order matters more than the tool. Ask AI for the idea itself and the work drifts towards what everyone else is getting, so the idea has to come first.

Where does AI help most in creative work?

AI helps most when a team needs to see more options than it could draw by hand. Sketching routes, testing a mark at every size and adapting one layout to 40 formats are where generative AI in design earns its place.

Research supports the exploring part. In a study with more than 800 participants designing virtual cars, AI-generated design galleries led to deeper engagement, longer exploration and better results (Swansea University, 2026). In an online experiment on short story writing, access to AI story ideas made stories more creative, better written and more enjoyable, especially for less creative writers (Doshi & Hauser, Science Advances, 2024).

Scaling is the other half. Once an identity is set, AI can produce the variants a brand needs across markets, channels and pack sizes. We see this as the clearest gain, because it gives people more time for the work that needs judgement.

Why should the idea come before AI?

AI pulls everyone towards similar answers, and a brand only works if it is different. The same Science Advances study found that stories written with AI ideas were more similar to each other than stories written by people alone. The authors call it a social dilemma: each writer is better off with AI, but together the writers produce a narrower range of new work.

For AI in branding, that narrowing is the main risk. Recognition rests on distinctive brand assets that nobody else owns. An idea sourced from the same models your competitors use starts from the same place theirs did.

The second risk shows up later in the project. In a yearlong study of senior creatives running TV and streaming campaigns, AI outputs, mostly made by one person with one tool, looked so finished that clients fixated on details that were impossible or costly to reproduce. Production teams then had to spend unplanned time reworking ideas and outputs (Harvard Business Review, 2026). The researchers advise being clear about what an AI-generated concept does and does not represent.

Conclusion: Bring the idea yourself, then let AI find its versions.

Should AI-made work meet a different quality bar?

No. The bar is the same whether a person or a model made the first draft: does the work fit the brand, and will it hold up in real use?

People lower the bar when they trust the tool. In a survey of 319 knowledge workers, higher confidence in generative AI went with less critical thinking, and higher self-confidence went with more (Microsoft Research and Carnegie Mellon, 2025). Respondents also reported less critical effort on routine, low-stakes tasks under time pressure. The findings are self-reported, but most everyday banners, variants and social assets are made under exactly those conditions.

Judgement is the part a model can't supply. When IW renews an established brand, every asset is assessed first, from logo and colours to typefaces and pack details. Each one is kept, refined or complemented with something new before anything is redrawn. In the Lurpak case, the task was to help a brand built on 120 years of craft heritage evolve and be ready for the future. A brief like that depends on knowing which details customers would miss.

Conclusion: A cheaper draft still has to clear the same bar.

Isn't AI already more creative than most people?

On some measures, yes, and this is the strongest argument for letting AI lead. A study described as the largest direct comparison so far between human and large language model creativity found that generative AI can now beat the average person on some creativity measures (Université de Montréal, 2026). On that reading, starting with the model looks like the efficient choice.

The same study points the other way at the top end. The most creative people still had a clear and consistent edge over the strongest models, and results depended heavily on how the instructions were written. A brand needs an idea that fits one company only, plus someone who can write the brief that gets there. The narrowing Doshi and Hauser found still applies to the field as a whole.

What does this mean for your brand?

• Write the idea down before anyone opens a tool. Say in one sentence what the work should make people feel and why it belongs to you. AI explores from there.

• Use AI where volume matters. Give it routes and variants. Keep the choice of which route wins with people.

• Say what a concept is. When you show AI-generated work, mark what is decided and what is placeholder, so no one gets attached to a detail production can't deliver.

• Keep one review standard. Judge AI-assisted work against the same brand criteria as everything else, using clear rules, references and review.

Frequently asked questions

Will AI replace designers?

AI will take over much of design production, but not the judgement about what a brand should look and feel like. We make the full argument in will AI replace design agencies.

What is generative AI in design?

Generative AI in design means using models that produce images, layouts, type or copy from a prompt. In brand work, it is most useful for exploring routes and producing variants of an identity that is already set.

Does AI make creative teams less creative?

There are early warning signs. An MIT Media Lab study found that people who wrote essays with an LLM showed the weakest brain connectivity and the lowest sense of ownership of their work (Kosmyna, Maes et al., 2025). The study is a preprint that has not been peer reviewed, and its sample is small, so treat it as a signal rather than proof.

Do AI-generated brand assets need to be labelled?

Some uses carry transparency requirements under Article 50 of the EU AI Act, which has applied since 2 August 2026. Our guide on labelling AI-generated content explains who is affected.

Where should AI sit in your creative process?

Keep the idea and the final judgement with people, and give AI the work in between, where options multiply and output scales. The quality bar stays where it was. If you want to talk about where AI fits in your brand work, write to us at hello@iwagency.com.

Sources

• Generative AI enhances individual creativity but reduces the collective diversity of novel content — Science Advances, 2024

• Research: Gen AI Is Collapsing Creative Processes — Harvard Business Review, 2026

• Scientists discover AI can make humans more creative — ScienceDaily / Swansea University, 2026

• Researchers tested AI against 100,000 humans on creativity — ScienceDaily / Université de Montréal, 2026

• The Impact of Generative AI on Critical Thinking — Microsoft Research, 2025

• Your Brain on ChatGPT — MIT Media Lab, 2025