Inside WPP’s AI brand hub—how marketers can experiment and what’s next

In early 2023, as ChatGPT was becoming all the rage, WPP set out to identify use cases for AI that would help its clients meet their creative and strategic demands. A year later, the advertising giant announced the culmination of its efforts: Open, a re-imagined technology hub backed by annual investments of $318 million and used by over 35,000 employees to service blue-chip clients including Coca-Cola, L’Oréal, Nestlé and Ford. Open reflects a growing trend among advertising companies, which are developing powerful AI platforms that connect disparate large language models (LLMs) for a variety of client needs, including experimental content generation, real-time commerce updates and faster campaign optimization. See all of Ad Age’s 2024 Creativity Award winners here All large players will eventually need to adopt this strategy, according to Stephan Pretorius, chief technology officer at WPP. Publicis, for example, unveiled CoreAI in January, fueled by a $326 million investment in AI technology over the next three years. But as advertising companies race to build out these platforms, a wave of concern about how AI is used is moving across the industry. Production companies are contending with creative threats and brands are asking agency partners for tighter guardrails around the technology. The AI craze persists, yet, some marketers are drawing lines in the sand. Also read: Google AI ad tool restrictions marketers should know To get a deeper look at how WPP is navigating these issues while seeking to provide clients with a robust AI sandbox, Ad Age sat down with Pretorius, who demoed Open’s primary offerings, dubbed “studios,” where the latest LLMs from providers such as OpenAI and Google are chained together for maximum effect. Inside a client workflow on Open. Credit: WPP (Click image enlarge) Content machines In Open’s “creative studio,” clients are offered a wide-reaching playground to experiment with generative AI tools. The space integrates models such as Gemini, Stable Diffusion and GPT-4, allowing brands to develop both text- and image-based creative, while also digesting various buckets of data—including that of the brand, WPP, licensed and public—to enable more relevant outputs. WPP is hoping to eventually add functionality for other generative abilities, such as voice for language translation, 3D modeling and video. The potential of text-to-video tools has recently grabbed advertisers’ attention thanks to OpenAI’s release of Sora. All of this opportunity comes the worry that data will get lost, or even hoovered by the models themselves. This concern has been a major sticking point in the relationship between AI companies and users and has fueled hesitation from governments and advertisers alike, who are worried that what they consider private or proprietary may no longer be so in the hands of this new technology. Also read: Why brands are adding AI restrictions to agency contracts The usefulness of Open’s creative studio, said Pretorius, is in its reassurance that clients’ data is protected. This security is one reason why brands are interested in AI sandboxes, provided by both ad companies and tech giants, for honing skills like AI prompting. Settling with free, public versions of platforms like ChatGPT is a sure-fire way of surrendering any input data to the underlying model. Like creative studio, “content generation studio” is where clients can develop simple media for hypothetical ad creative, such as mock email and blog headlines, or mood boards illustrating different pieces of content based on different combinations of audiences and occasions. The studio’s focus, though, is offering a glimpse into how AI models can be used together to form a whole greater than the sum of their parts. One such outcome is what WPP refers to as a “professor of your brand.” The idea is that a client can feed their most relevant materials, such as a style guide or product descriptions, to various AI models, creating an assistant that knows them inside and out. This bot can then be consulted as a brand expert at numerous stages in the creative and strategy development processes. Moreover, when models and chains of models produce outputs in the content generation studio, the client can ask those tools to explain precisely how they arrived at their output. This transparency is meant to strengthen trust between brands and AI tools—perhaps the ad industry’s most recurring source of friction when it comes to AI, thanks to rogue chatbots and biases adopted through training data. Another way that WPP is squashing concerns is by preventing work produced in both creative and content generation studios from being used in public-facing activations. Some may see this policy as a limitation, but it is consistent with how other marketers view the readiness of generative AI. That said, another studio in Open is already primed for immediate commercial application. Where AI meets the real world Open’s “commerce studio,” which generates updates for retail sites, seeks to address an issue that brand marketers know well, said Pretorius. “Physically, you cannot maintain a million pages of content and [expect] them to be good,” he said. For this reason, Open connects different AI bots—each powered by a specific group of data—to maintain these pages for a client. To generate updates for their commerce sites, a client such as L’Oréal selects a series of criteria, including tone of voice, market, retailers and shopper barriers. The bots then tie these references together to form a targeted update, which if deemed acceptable, is then quickly brought online to the website itself. Related: Dove pledges not to use AI to portray real people in its ads While the work developed here can be used commercially, human intervention still plays an important role, said Pretorius. Employees are constantly keeping tabs on what the AI is generating, and a review system is in place for content that is on its way to the actual sites. Human replacement is a concern that is top of mind for the ad industry: Over 30,000 agency jobs are expected to be automated by 2030, according to Forrester. Perhaps even more insidious has been the fear that AI is becoming so advanced, humans won’t be needed to oversee it. But Pretorious is confident that won’t happen: “We are so far away from [AI] generating and distributing autonomously,” he said. Another studio with real-world applications is “media studio,” which is helping clients diminish media waste by streamlining optimization for video and static ads. It works via an AI bot that can predict how different features, such as text overlays, messaging and product depictions, will affect brand lift. Formerly, media agencies needed to wait several days after a campaign launched to analyze these kinds of results, losing money and time in the process as a consequence of running underperforming creative, said Pretorius. But Open’s predictive AI allows clients to test their campaigns before they run or edit them. “The benefit of [this] is enormous because you effectively can tell a client [they] will never make bad ads live,” he said. Managing the noise The inverse of brands’ concerns about AI technology is their delusions about its capabilities, which WPP is having to manage as it brings more clients into workspaces on Open. “People extrapolate things,” Pretorius said, explaining how a client in a recent meeting voiced the incorrect assumption that AI could generate complete ads for social media, without any steps in between. What’s needed in these moments is a shift in attention to what AI can do. For example, some clients working with WPP have only just discovered that generative models are unable to create brand-specific logos and packaging, said Pretorius. Their renditions of these details—which are some of the most critical in any brand’s ads—simply do not reflect flawless realism. But where AI can help is in developing the surrounding environment in which brands’ products are placed. This is where WPP’s partnership with Nvidia has come into play. WPP has worked with clients, including automotive brands, to generate backgrounds for ads, saving both costs and time. Nvidia’s technology streams high-quality content to customers, allowing for 3D rendering of diverse landscapes. “It’s like gaming tech gone completely enterprise,” said Pretorius. Over the next few years, WPP is hoping to automate the automation itself—that is, revamping the interfaces with which clients use Open’s tools. Out will go siloed dashboards that signpost when a client is exiting and entering disparate studios, and in will come a more conversational, cohesive workflow driven by what else but AI bots. “Clients are interested in distinctions, they’re interested in applications,” said Pretorius.

Inside WPP’s AI brand hub—how marketers can experiment and what’s next

Featured Stories