Two truths and a lie about Generative AI
Learnings from creative teams who are leading the way
Learnings from creative teams who are leading the way
In today's rapidly evolving market, creative teams face unprecedented pressure to deliver at the speed of culture, localise content and personalise experiences.
AI can help address these challenges, but not without expertise and strategic implementation.
In a recent webinar, creative innovation experts from WPP and Stability AI, operating at the forefront of AI adoption, shared their insights on the common roadblocks, critical breakthroughs and effective processes that are working for brands.
Read on for the key takeaways, and you can also watch the full webinar here .
The lie: AI is an easy button for creative challenges
Perhaps the most common misconception about gen AI is that it's a simple solution to complex creative challenges. The reality is that AI is not a magic button for cranking out campaign-ready content.
As Ana Guillén, VP of Marketing and Comms at Stability AI, put it: “While we all know that AI is here to help, not all AI is actually helping right now.”
AI makes it easy to produce average work, not great work. Which is why creative expertise is more valuable than ever in getting the outputs right – and also why it’s important to partner with AI experts who can get the most out of the technology.
“You can take what is quite good work to be really great work with AI, really enhancing the workflow, knowing the powers of different models and how they work and when to use them,” said Stuart Foy, Senior Creative Innovation Designer of WPP agency T&P.
Truth #1: Out-of-the-box AI doesn't know your brand. You need customisation.
AI models don’t come trained on your brand’s visual identity, which means the results don’t look like your brand: They look like everyone else’s.
As one client told Stability AI: "If we wanted to be like everyone else, we would stick with the other tools."
This is why customisation is critical. Fine-tuning, the process of training AI models on brand-specific datasets, enables professional creatives to achieve the quality and consistency their brands demand.
Dennis Niedworok, who leads Stability AI's applied research team, explained the process: "We take the brand playbook, digest it, and turn all of the important things from it – like the photography style, the general look, and the colors – into a fine-tune. We show our model all of these images and tell the model, 'This is what we want to see.'"
The process is highly iterative and hands-on, requiring multiple rounds of collaboration to ensure every detail aligns with brand standards. “It's really important to have real in-depth knowledge of your brand and be able to see these things and know what's good and what's right,” Foy said.
The result? Campaign-ready assets that maintain brand integrity down to the finest details.
Foy saw those results after fine-tuning a model for client Argos the beloved UK brand.
"The AI model picked up more than just the characters. It picked up the really fine grain that we used in our 3D animations. It picked up the lighting, and subtle shadows that I wasn't even thinking about."
Truth #2: AI has a UI problem. Brands need bespoke platforms
Powerful AI technology remains largely untapped if creative teams cannot access it effectively.
Months ago when Stability AI started offering custom enterprise solutions, the team delivered an AI workflow to a major mobile manufacturer and the response was, "This is incredible. How do we use it?"
“It was an eye-opening moment for us. We needed to have a front end that makes this really complex technology very easy to use,” Guillén said.
AI technology is useless if creative teams can't access it effectively. But many UIs are “disconnected, complex and confusing,” as Foy put it. Creative teams are often stringing together multiple platforms and models, juggling subscriptions and moving assets from place to place.
In response, clients are working with WPP and Stability AI on custom UIs built specifically for brands. Dan Northcote-Smith, Creative Innovation Director at T&P, appreciates the partnership with technical experts to build accessible tools that enable him to create with speed.
“They've got these complex workflows in the backend, but then they're going in them and making them accessible to everyone. They're designed for creative production,” Northcote-Smith said.
WPP Open and Stability AI’s Dream Studio are both examples of customisable UIs built for creatives. Watch the webinar to see a custom UI in action.
Real-world success: How Argos moves faster with custom character models
T&P worked with retailer Argos to launch a digitally-led campaign featuring two popular 3D toy characters, Trevor (a small dinosaur) and Connie (a fashionable doll). The challenge: Enabling those characters to appear everywhere the brand needs to be, whenever opportunity strikes.
Traditional 3D animation doesn’t deliver reactive content fast enough. Responding to a major event, like a Formula 1 race weekend, requires immediate content generation, not a weeks-long animation cycle.
So WPP’s T&P worked with Argos to train custom AI models on the characters. As Foy described, the training wasn’t just about getting the characters to look right, but also how the characters interact with the world. The models learned Trevor the dinosaur’s unique proportions, understood that he couldn't hold objects like a human would, and even developed a “meta understanding" that allowed the characters to interact with plush versions of themselves.
The results were high-quality images that looked like screenshots from the TV adverts, generated in minutes instead of months. The custom AI solution maintained the characters' beloved personalities while enabling the brand to create seasonal content, react to cultural moments, and produce variations at scale.
The path forward for creative teams
As the webinar made clear, successfully implementing generative AI in creative production requires three things:
The promise of AI in creative production is real, but realising that promise requires moving beyond out-of-the-box tools and instead toward custom production platforms powered by highly iterative, fine-tuned models
The tools exist. Now brands need to do what they do best, and use them creatively.
As Northcote-Smith said, "We are living in the most amazing time to be a maker who has tons of ideas and just wants to get them out in the world."
To partner with WPP and Stability AI to build custom AI that works for your brand, reach out to kathy.ponting@wpp.com, or learn more about the Brand Style solution here.