From Dentsu to Apollo Partners, the buildout of AI tools and talent at agencies is picking up pace

The tools are designed to enhance Carat’s (and other Dentsu agencies’) media and creative strategies by creating consumer personas, assistants for brainstorming and automating trend analysis. Ingesting a variety of data formats from Dentsu’s own research (such as its Merkury data center) as well as clients’ first-party and other third-party data sources, the main goals are to dig deeper into insights and faster decision-making, with the side result being saving time and cost. 

As Michael Liu, evp and head of innovation at Carat, put it, despite the potential for AI to reduce headcount and save time, the Vurvey partnership actually helps to supplement teams with another virtual brain — one that thinks fast and intuitively once it’s been taught properly.

“We can build assistants, which are sort of like myself and other of my colleagues, specialists who are dedicated to trend spotting, or understanding insights — someone who’s dedicated to brainstorm with us,” said Liu. 

(He immediately clarified that the goal is not to replace colleagues but to add to them. “These are colleagues that we have 24/7 across our global teams,” he said.)

“We build agents that are more reflective of real people, as opposed to chat bots and those kinds of things,” added Chad Reynolds, founder and CEO of Vurvey Labs. 

“It’s not just bolting on a sexy UI on top of open AI. It’s their model that’s based off of understanding people. And that’s that’s inherent to what Carat is about too,” added Liu. 

As with an increasing number of AI platforms, unstructured data can be restructured and doesn’t need prepping before being entered, be it spreadsheets, PDFs, decks, consumer profile reports, research reports, earnings, reports or competitive Intel — a massive time saver, said Liu.

In a test project with an unnamed restaurant chain, Carat and Vurvey dug into understanding casual dining behaviors, to go beyond what can be gleaned out of quantitative data.

Because of Vurvey’s massive warehouse of video respondent information, crunched by AI, which Carat tapped into, Liu and team were able to not only goose the media strategy but also offer input on the creative side of making the ads. “It took me longer, to be honest, to write the questions that I wanted to include in this test than to get responses back,” said Liu. “Because we have access to these agent personas, they can let us know whether it’s social or out of home or TV or CTV [that resonates best], what types of formats and what types of things catch consumers’ eyes as well.”

The arms race for AI development isn’t limited to holding companies, as independents large and small build out their expertise and tools. Apollo Partners, a 20-30 person media agency founded by Eric Perko, opted to hire a new hybrid type of personnel to helm such efforts — executives who blend media skills with coding abilities. 

Digiday has learned that Apollo has brought aboard Trevis Milton to head up automation and AI solutions. Milton has media agency experience from time spent at Starcom, but he most recently used hands-on coding skills he picked up along his career path to apply automation and AI solutions directly to media workflows at Yahoo. 

Perko and Milton explained their aim is to build AI solutions and tools from the ground up, in an attempt to give rank-and-file planners exact tools, including an internal quality assurance process on campaigns that reduces the tire-kicking efforts of QA from eight hours to two minutes, said Milton (which he added, will become fully agentic within a year or two). 

“We are putting systems and process in terms of automation and AI into the hands of the people who are doing the work every day, the planners, the buyers, the supervisors who have, unfortunately, a pretty heavy manual process when it comes to certain parts of the media planning and buying process,” said Milton. “Anytime you have that, you have error prone opportunities for risk. And so we’re starting with that.”

Not that partnerships with other tech providers won’t be pursued, added Milton. But the driving goal is to reduce risk and increase deterministic outcomes. “In order to evaluate a build that we’re going to do or a partnership that we’re considering, we have to start with our own strategy,” said Milton. “We want to start with deterministic automation first. That means that we’re building where we have the absolute highest degree of confidence in the outputs we’re generating.”

“There’s plenty of folks out there who know how to code around AI, but don’t know how the media business works,” added Perko about Milton’s hire. “Travis has been here for two weeks, and we’ve already started to accomplish these things.”

At this early point in AI’s advancement across the agency landscape, it doesn’t appear the end goal is to replace humans. “Part of what we hear from the CMO is a narrative that is designed to quell some of the concerns employees might have about having their job being automated away,” said Jay Pattisall, vp and senior agency analyst at Forrester.

Still, organic attrition within agencies can sometimes be solved by using personas and intelligent tools like what are described in this story. It’s definitely something to look out for as AI creeps deeper into agency workings.