AI agents are the current focal point for agencies, marketers and publishers hoping to deploy the potential of generative AI to practical effect.
Examples range from Coca-Cola’s Fizzion project, an agent embedded into the Adobe Creative Cloud, which enforces the brand’s visual style guidelines across its global creative output, to tabloid The Sun’s planned programmatic media agent.
Additionally, a Gartner survey of 400 martech execs published this week found 81% were engaged in a pilot or rollout of an AI agent at their company. Use cases differ across business verticals, but the underlying ambition is the same: eliminate grunt work and create economies of scale.
With interest running high, a number of key industry players recently allied to develop AdCP, an open-source communication protocol that lets AI agents — whether built by advertisers, publishers, or ad tech intermediaries — interact using a common language.
To get the lay of the land, Digiday spoke to a clutch of media and creative agencies, publishers and platforms about their progress:
- Agents and agentic AI have caught the industry’s imagination, but are mostly limited to advisory and information-retrieval roles.
- Specialist media agents are backseat drivers, advising rather than placing spend while supply-side agents help consolidate data and assist execs.
- Economies of scale are within reach, but limited by older working habits. Meanwhile, quality and dependability of agents varies considerably; agency staff find themselves working with both their own systems and those of their clients.
LG Ad Solutions (one of the founding members of the AdCP protocol) has created a platform designed to coordinate its own agents, and those belonging to other advertisers, agencies or adtech players called Agentiv.
Dave Rudnick, chief technology officer at LG Ad Solutions, said the system’s media effectiveness agent had cut the time taken to compile campaign reports from two days to five hours. None of the agents LG had designed would handle money, he explained.
“Today, they’re making recommendations. We want to be able to allow the experts to continue to control the knobs,” Rudnick added.
Having built the system, LG Ads is now talking to agency holding companies and brands about plugging it into their own nascent agentic systems. (A spokesperson for the company declined to name the companies it’s speaking with.) Rudnick said the system, which will include up to 30 distinct agents, should enable LG Ads to offer greater efficiency to advertisers buying through its DSP – making agentic AI the latest front in the perpetual competition among DSP providers.
It’s still early days, however, and it’s unclear whether agentic media buying or publishing agents can actually deliver today or not.
At indie media agency Mediaassociates, staffers have been using proprietary AI agents for the past year. Tim Lathrop, vp of platform digital, told Digiday the company had created agents to assist with media planning work like making audience-segment suggestions based on client briefs, running competitor analyses, and suggesting potential keywords a buyer might bid on in a search environment.
Once again, none of the agents actually execute media buys; spending decisions remain the domain of real people. “The money side will be more of a long-term process,” said Lathrop. “Now, it’s [about] how we can build efficiencies around media planning and troubleshooting.”
On the other side of the aisle, publishers are exploring the use of agents for both media and content ends. Some are farther along than others. Mario Lamaa, managing director of data and revenue operations at Immediate Media, said the publisher has built an AI-powered agent that sits on top of its PRISM first-party data platform, giving sales teams instant access to information that once tools days to pull.
The tool pulls together audience and contextual segments, campaign performance and historical post-campaign analysis, allowing reps to respond to client briefs in real time instead of waiting on analyst reports. Lamaa said immediate is now working on a second version of the agent that will integrate with additional data sources, particularly research and synthetic insights — to make its recommendations more comprehensive.
“Our sales team can query [the AI agent] whenever they are in conversations to help them respond to briefs,” Lamaa said onstage at Digiday Publishing Summit Europe conference in Lisbon, Portugal, on Wednesday (Oct. 29).
“If we were speaking to a car client, they could speak to the agent and ask it: ‘what kind of segments do we have around this? What size are they? How could we deliver a campaign over a month?’ That’s the stuff that typically we would have to ask one of our analysts…. [with] a turnaround time of around 48 hours,” he said. “It means that we can be responding to our clients … in a much shorter time than we ever could in the past.”
And Rob Lang, newsroom AI editor at Reuters, said he saw an opportunity to use an AI agent to assist with time-consuming video editing tasks, such as cutting together clips that take into consideration continuity errors like a subject taking their glasses on and off.
“We might be able to get the AI to build the edits for us, which would be quite extraordinary,” Lang said on stage Tuesday (Oct. 28).
Creative agencies’ use of AI agents also focuses on lancing the mundane. At Rapp, for example, staffers have access to a variety of agents through parent company Omnicom’s Omni operating platform. James Heimers, evp, marketing sciences at Rapp, told Digiday that an agent used to pull historical data from within Rapp’s own client database was now the “starting point” for presentations and sales decks. Such systems have sped up creative strategy work, he said, but obstacles remain.
“For us to get from a brief to content, whether that’s imagery or copy, that can happen in a day now, on a process that was probably taking three or four weeks. The big blocker is approval processes [that] haven’t caught up,” Heimers added. “You’re probably saving 50% of your time … that could be substantially higher if we had better processes around the edges.”
At creative agency Mekanism, co-founder and ECD Ian Kovalik said staffers have been using agents for six months. Using agents based on Gemini, Google’s LLM, for work such as video editing, Kovalik said the average time taken to shoot a video had fallen from eight weeks to one. The agency had cut outgoings on production and VFX companies, too, though Kovalik declined to estimate the cost savings.
“It’s like having instant production at your fingertips,” he said. The company had also begun using an agent designed to trawl through its previous pitches and campaigns, and collate relevant information for staffers filling out requests for proposals (RFPs). “It’s a huge time saver,” he said.
Like Heimers, Kovalik recognizes the imperfections in agentic work. “We’re still experimenting,” he concluded.