• by Laurie Sullivan , Staff Writer @lauriesullivan, March 31, 2025

AI will take on a larger role in the entire advertising process -- from building creative and media buying to optimization and running reports.

Jellyfish -- which has been rebranded with a new look and experience -- recently announced the use of AI agents to buy media across Google, Meta, and Amazon. The technology has begun to carry out tasks on behalf of people, including processes from automating campaign setups to optimizing performance.

These agents not only buy media, but also adjust budgets daily to avoid under- or overspending, organize campaign data, and set up reporting dashboards. Tasks that once took weeks and demanded hundreds of hours of human oversight can now be accomplished in much less time.

“We looked at the media process to determine where people spend data-entry time on work that isn’t strategic or creative that no one really wants to do,” said Jeff Matisoff, partner at Jellyfish. “It’s not like they say, oh, I can’t wait to set up 1,000 IOs in Google Display & Video 360.”

Many of the 65 engineers are well-versed in AI. Developers at Jellyfish built agents that can work across the taxonomy. The agent can go back into DV360 or into Meta’s platform, set the audiences or fix the line item and correct the content, and set optimization. Humans check the changes to ensure nothing goes wrong.

Once an agent is trained to work in a platform and knows what Jellyfish reps and the client understand as being the best ways to run the campaign for key performance indicators and budget, it can optimize the campaign to get the best KPI and ROI.

“We’ve seen some really fast optimization techniques for clients,” he said.

The technology, which updates every 15 minutes, catches mistakes and fixes them in real-time, and then runs a report. It can mimic the process and decision that a junior-level media buyer can carry out.

Once AI knows the rules set by the brand and agency it can identify and self-correct any mistakes made to run the campaign. For example, if someone entered a bad country code such as USK rather than U.S. or UK to run the ads, the AI would identify and correct it.

“Reporting is always accurate,” Matisoff said. “We don’t have a situation where we’re missing $4 million in the latest spend because the taxonomy didn’t map. The agent gets trained on how a specific client works.”

Some might think automation will reduce the need for large teams, but Matisoff doesn’t see it that way.

While it takes less time to launch a campaign by 65%, due to the speed with which the tech performs various tasks, the agency continues to hire employees to support the more than $2 million in media budgets it handles.

Jellyfish also wanted to ensure that once campaigns are set up on a specific taxonomy, reporting is trusted and correct.

The key is perfecting media mix modeling (MMM) output to look at the perfect ways to allocate budgets based on a brand’s real performance. Matisoff said the company is able to get to near real-time data to understand how to push budgets across channels and countries with products.

Jellyfish became one of the first certified users to support Google’s AI for MMM, Meridiam.

The agency also has social AI agents that scan Instagram, primarily, for insights on how other competitors communicate across organic, paid and social content, providing insights into strategy and creative producers to offset the competition.

One AI agent called Share of Model, can provide search engine optimization (SEO) but for large language models. In the same way SEO helps marketing and search engine, the agent looks at the relevancy of content and recommendations to identify why.