Agency holding groups, including WPP’s Ogilvy, IPG, Omnicom Group and Havas, are taking a more strategic approach to how they use generative artificial intelligence internally and with clients, a further indication of the industry getting to grips with the burgeoning tech.
Over the last six months, more agencies have formed cross-department AI task forces to navigate the opportunities and limitations of generative AI. In some cases, these have led to winning business and forging enterprise partnerships.
Ogilvy Paris won two clients in the last two weeks thanks to its commitments to building its AI Lab, an internal committee dedicated to driving AI expertise so brands can create impactful campaigns, said president Mathieu Plassard, who wouldn’t share client specifics.
Omnicom’s AI policy led to its enterprise partnership with Microsoft which was announced in June, according to Paolo Yuvienco, evp and CTO, Omnicom. The partnership gives agency staffers access to OpenAI’s GPT models through Microsoft Azure.
Agencies continue to use generative AI tools to help save time and lower internal production costs. Still, there is caution with experimentation, especially since some client partners are yet to embrace the tech, given the growing list of potential harms, such as AI-produced factual inaccuracies, IP infringement, and ownership concerns.
“[These policies] are helping agencies [under IPG] understand how to put structure against using this technology and ensuring the safety of our clients and their brands,” said Jason Snyder, CTO, of IPG’s Momentum.
Inside the agency policies
At Havas, its Play, Don’t Push AI policy took shape in January this year and lets employees use generative AI but with some caution. Employees are not permitted to use the content output verbatim for any real-world campaigns.
“These are sensible procedures anyone will observe,” said Mark Whelan, chairman and group CCO of Havas UK.
The policy also requires Havas to be transparent with its clients anytime it uses generative AI in campaign ideation or for workflow, such as strategy ideas via ChatGPT or concept visualization through Midjourney, a tool that creates images based on text prompts.
Omnicom’s AI policy helped land its partnership with Microsoft, integrating OpenAI’s GPT models for the agency’s Omni Assist tool, which works as a virtual assistant helping with tasks like audience insight and campaign planning. However, employees aren’t permitted to present work solely produced by generative AI to clients, unless it’s a direct partnership on specific AI initiatives, said Yuvienco.
Meanwhile, IPG’s generative AI blueprint is used by Momentum for internal operational tasks, such as generating social media content or blog posts. Its policy also includes the use of AI for client needs, such as analyzing customer data to produce relevant content like emails or targeted ads. All AI-generated content undergoes a rigorous review process before it is published or shared, the agency said.
The policy states that the client assumes brand safety risks, including IP infringement, related to AI tech. Even so, the generative AI tool has been trained to ensure it does not infringe upon any third-party intellectual property, said Snyder. The agency is required to anonymize all data to protect the privacy of its clients and their customers
Momentum’s marketing guideline highlights the ethical use of AI, including provisions on prohibiting deep fake content.
In June, Ogilvy introduced the AI Accountability Act, which aims to enforce transparency for brands using generative AI-enhanced influencer content by including the hashtag #poweredbyAI.
Ogilvy has also implemented internal policies, such as ensuring human oversight of creative outputs and adhering to client confidentiality when inputting data into the large language model (LLM).
Anatomy of the internal task force
At Omnicom, the cross-department task force consists of agency leaders, including members of legal, privacy, data, creative and user experience teams. The group is tasked with driving the practical applications of generative AI.
“The scope is to have early access to all the tool sets made available through the partnerships,” said Yuvienco. “And then to ensure that we’re applying the appropriate legal and ethical eye towards it.”
Havas built its task force in January at a global and agency level. The group meets weekly and includes members from operations, legal counsel, and frontline client executives to give a centralized perspective regarding compliance matters, innovation, platforms, and partnerships.
Elsewhere, Ogilvy’s AI Lab meets weekly to review projects and conduct workshops with new tools, and includes senior client service employees, an internal legal and finance team, and a creative tech team across WPP, said Plassard.
Relevant ads in real time
Momentum is exploring combining different types of generative AI models, such as text-generating, image-generating, and music-generating models to come up with solutions for clients. These range from creative ideation to content production and campaign execution said Snyder. It is also exploring how to address environmental and diversity issues with technology.
In the next 36 to 60 months, Omnicom is aiming to achieve relevant ads in real-time for brands through generative AI, a timeline that CEO John Wren has previously discussed.
The agency employs templating for automation, which requires time to develop to the point where they can be scaled across different channels, said Yuvienco. Once the templates are in place, generating variations becomes much more simple.
Currently, using generative AI to create relevant ads involves significant technical infrastructure and computing power.
“This is why it is so important to have the first-mover advantage with the major tech and cloud players,” Yuvienco said.
