At the Cannes Lions International Festival of Creativity in June, marketing leader Les Binet had a conversation with himself, or rather a ChatGPT-powered version of himself. Binet, Adam&eveDDB’s global head of effectiveness, demonstrated how the bot, dubbed LesGPT, can provide marketing advice after being trained on his decades of knowledge and experience in the advertising industry. The AI speaks in Binet’s voice, too.

LesGPT is one of nearly a dozen bespoke AI tools developed by ad agency DDB Worldwide to help marketers shape briefs into usable campaign inspiration. Adam&eveDDB is an agency within Omnicom Group’s DDB Worldwide.

Like many major agency holding companies, Omnicom Group is pursuing AI and deploying the technology throughout the organization. But generative AI, which includes the creation of text, image, voice and video content, is still very much in its infancy, and agencies such as DDB Worldwide are looking for breakthroughs that could help their teams and brands. In fact, Binet is not the first marketing leader to be embodied by generative AI—Publicis Groupe made an AI replica of CEO Arthur Sadoun at the end of last year to make 100,000 personalized videos for the holding company’s staff. And at Cannes this year, Monks created a robotic alter ego of its owner Martin Sorrell. 

Check out our AI marketing glossary 

The ad industry’s need for practical AI is growing as CMOs struggle to advance use cases beyond the lowest-hanging fruit in their marketing plans. Despite massive investment in the space, AI-powered wins have remained elusive, making experiments like those from DDB a useful exercise to push marketers to think about AI in new ways. 

That's why Ad Age recently sat down with George Strakhov, DDB Worldwide’s global head of creative technology, who demonstrated five of the agency’s custom AI tools. Over the last 18 months, Strakhov has led the development of DDB’s AI funhouse, which is part of its tech-focused creative platform called RAND. Strakhov, who is based in Amsterdam, has been taking this new toolbox on the road to educate employees around the world on the new opportunities. Approximately 10% of DDB’s workforce of 6,000 people are now weekly active users of these tools.  

As AI becomes omnipresent in the world of advertising, Strakhov said he is finding more reasons to push beyond boring use cases. “For our industry, it’s really important to kick against the gravity of the cliché,” Strakhov said.

DoneOrNot weeds out tired creative

There is a theory that every original idea has already been conceived, which, if true, would be bad news for creative marketers. DoneOrNot aims to give a straightforward answer to this concern by searching the web for a campaign idea to see if it’s previously been carried out.

The user inputs a brief such as: “Campaign idea for a car brand, where we shoot the commercial in real time and broadcast it live.” DoneOrNot deconstructs the brief and scours the web for its components. The AI will specifically search for whether the idea has been implemented as part of an ad campaign.

After about a minute or so, DoneOrNot reveals its findings. It pairs each example of a historical campaign it finds with a number between one and 10, corresponding to how similar the campaign is to the idea inputted (10 being the most similar). For example, for the brief mentioned above, DoneOrNot found that Nissan produced a four-hour livestream with a creator on YouTube in 2023. (Some of the outputs, however, did not feature commercials shot and broadcast in real-time.)

These insights help creatives decide more quickly whether to follow an idea, Strakhov said. But even if they find that a brief has already been executed, DoneOrNot will suggest ways to spin it a bit differently. The result could allow for a unique twist on an otherwise dead end.

Related: Brands add AI restrictions to agency contract

LesGPT was trained on marketer's work

Binet is a valuable asset for DDB clients—he’s an econometrician, has degrees in physics and cognitive science and has been working with AI since the 1980s. 

“But the bottleneck is we only have one of him,” Strakhov said.

That was the motive behind LesGPT, the chatbot trained to be Binet’s digital replica, available for client work regardless of where the real Binet is located in the world. The knowledge held by LesGPT has been distilled from Binet’s published and unpublished research.

Not only does the bot provide marketing wisdom at a depth that normal chatbots may struggle to reflect, but it can do so in Binet’s voice. For now, LesGPT can only output text and dictated text, but soon it will be able to produce charts to better illustrate its points, Strakhov said.

Freak Factory lets the facts fly

Freak Factory is designed to help marketers think about a challenge, such as a campaign or a business’s value proposition, from a different angle so that creative solutions can arise, Strakhov said. The idea is to leverage the power of pithy facts—“facts that will make [someone’s] eyes pop,” Strakhov said—to galvanize this process.

A user inputs a brief, no longer than a sentence, such as: “We need to help teenagers battle screen addiction.” An AI system then scrapes its training data—mostly Wikipedia entries and YouTube videos—for relevant figures and statistics, Strakhov said. That information is passed to a second AI system to be formulated into a series of single-sentence, digestible facts.

Finally, these are displayed to the user along with their corresponding sources, so that facts can be verified and further investigated for creative fodder. 

BadIdeasBot gives cliché answers on purpose

Like Freak Factory, BadIdeasBot is about sourcing a multitude of ideas for the purpose of creative inspiration. The only difference is that the “bad ideas” bot is meant to spit out terrible campaign suggestions, hence its name.

BadIdeasBot sacrifices quality for quantity, which is core to its usefulness, Strakhov said. The bot takes a brief and generates obvious creative ideas. The AI utilizes little to no information about the target audience or key performance indicators, which makes its outputs broad and stale. This bot's suggestions also tend to lean heavily on puns.

For example, a brief that aims to encourage mobile gamers to switch to a new network provider outputs ideas such as “Lag No More,” which suggests—with little depth—that marketers show gamers the difference between an old and new provider. Another intentionally empty recommendation is “Rage Quit Your Network,” which entails empowering gamers to abandon their old networks in typical gaming style.

Discussing what doesn’t work helps marketers zero in on what does, Strakhov said. It’s analogous to how some writers pen a “vomit draft,” or a stream of consciousness dump of every thought that is on their mind in order to get those thoughts out. 

Strakhov does not rule out the possibility that BadIdeasBot could generate a good idea. It’s possible that a user comes across something that sounds pretty good, or at least the idea takes their mind in a new direction, Strakhov said.

EmailAI fits into the workflow

One roadblock to deploying AI at work is that accessing the tools often requires a “break in the flow” of someone’s usual way of doing their job, Strakhov said. For example, if a marketer wants certain files to be summarized by generative AI, they often need to exit a program on the computer and go to a new one. Small points of friction like this ultimately hurt adoption, Strakhov said.

To get over that hurdle, DDB created EmailAI so people could access the AI model as easily as reaching out to a coworker, Strakhov said. Whatever task the user has in mind, that worker can describe it in an email and send it to the AI, which handles the request and replies by email with its response.

The solution is simple, but it could make a big difference in getting workers over any initial resistance to working with AI, Strakhov said. If AI is made easier, that could shape how the marketers feel about AI in the workplace going forward, Strakhov said.