Inside Publicis Groupe’s AI strategy

When Publicis Groupe introduced its artificial intelligence platform 'Marcel' in 2018, much of the advertising industry was skeptical. 

Even Scott Hagedorn, now chief solutions architect at Publicis, admitted during a fireside chat last week at MM+M, PR Week and Campaign US' 'AI Deciphered: Discovering the Potential for Marketers and Communicators' event, that he was “digging in and laughing at the same time.”

But these days, everyone in marketing is racing to adopt AI—and Publicis has to continue to evolve its business and the way it works with clients to keep up. 

Hagedorn joined Publicis shortly before ChatGPT launched in 2022 to develop new solutions for clients that leverage major technology transformations. He had spent nearly two decades prior at rival holding company Omnicom running various tech and media businesses, and was involved in the development of its operating system, Omni. 

(L-R) Campaign US' Alison Weissbrot, and Publicis Group's Scott Hagedorn. (Photo credit: Erica Berger)

Upon arriving at Publicis, Hagedorn said he brought together cross-functional teams and tasked them with designing vertical-specific AI applications for audience building, media activation, creative and creative review.

A larger goal was to “automate the creative working process, which in some categories, like pharma, can be highly, highly manual and take a lot of time,” Hagedorn said.

With Epsilon and Sapient under its umbrella, Publicis has various pieces of technology that Hagedorn and his team are threading together, using AI, to create more efficient and effective media plans, such as Epsilon’s clean room and customer data platform.

“That is sort of fundamental, we think, to building AI—because if you don't have the data, and you can't train the models on great data, on an infrastructure—you sort of have nothing,” Hagedorn said.

Despite the potential of AI for clients, Hagedorn is acutely aware that AI will force agencies to need to adapt their business models even further away from time and materials.

“We're evolving into building a lot more algorithmic software, so we have to figure out how to monetise that,” Hagedorn said. One possibility is a software-as-a-service model, where Publicis builds and licenses tech to its clients, he added. 

Agencies and holding companies will also continue to gain access to the same tools, making differentiation on AI a challenge. According to Hagedorn, companies can stand out in the way they train AI models. 

“It will be interesting to see how everyone trains the models that they get in the open market and whether or not the finished product looks anything like what you start with,” Hagedorn said. “Everyone could go in radically different directions.”