When R/GA first rolled out its dedicated artificial intelligence practice, Brand AI, last July, the global agency touted the internal unit’s potential to produce branded chatbots, voice apps and other projects that put new AI-related technologies front and center.
A little over a year later, the division’s role has changed considerably, according to R/GA chief technology officer Nick Coronges. While it has certainly created campaigns with explicit AI elements—a chatbot for the Cosmopolitan of Las Vegas, for instance—the agency has found the tech to be of much more use in the behind-the-scenes workflow of its operations across the board.
“It’s changed a lot,” Coronges said. “While the original launch made a lot of sense to get the agency rallied around this technology quickly … our mission now is really to have data strategists and engineers plugged into all the work we do, rather than be a separate door into the agency.”
This evolution mirrors a broader shift in how ad agencies are treating AI and data science. As recent breakthroughs in machine learning have galvanized the business world, agencies like Epsilon, Heat, R/GA and Wunderman have formed internal units in the past couple years dedicated to integrating newly available AI tech into campaigns.
If it hadn’t already at launch, the work of many of those divisions has begun to permeate wider operations in their respective agencies to the point where AI is more of a commodity than a flashy accessory.
“I want AI to be this kind of invisible thread that runs through everything we make and do,” said Kathryn Webb, head of AKQA’s AI operations. “Imagine if we talked about the internet the same way we talk about AI. I want to get to a point where it’s a given, like we don’t talk about the fact that we use the internet at work every day.”
In R/GA’s case, the agency has found AI particularly useful for more rote applications like machine learning-powered decision trees and predictive analytics tools that can hone the user experiences of digital products.
“Most projects where AI techniques were effective weren’t designated as AI projects from the outset,” Coronges said. “If you start with AI, it puts people in the mindset of fitting a tech solution to a problem that doesn’t really exist. Instead, we use AI techniques to solve real problems we’ve identified.”
San Francisco-based Heat became the latest agency to unveil a dedicated AI practice in June of this year. The co-heads of the unit said it will focus mostly on back-end applications, like surfacing creative trends and personalization using services acquired or developed by parent consultancy Deloitte Digital.
“We don’t operate as a separate entity within the agency,” said Jocelyn Lee, co-head of Heat’s AI practice. “We are teaching and learning. And all of our different creative teams are using AI to help enhance the things they are doing so that we have a leg up with predictive insights and not just latent insights.”
In one project for a major shoe brand that the agency said it couldn’t name publicly, for instance, the unit used engagement and clickthrough data to tap into simmering trends tangential to sneakerhead culture in areas like gaming and tech. The creative team then used those conclusions to produce online ads that exceeded the client’s goal of 30% net new audiences within 30 days, according to Lee.
“This shoe retailer felt like they didn’t have the brand permission to speak to people in the way that would make them feel like they were relevant,” Lee said. “It really allows us to place bets on things we know are going to be huge.”
Software that makes machine-learning tech easier to navigate for the noncoding layperson has also contributed to its proliferation across agency departments, according to Ian Beacraft, vp and group director of digital strategy and creative technology at Epsilon.
“This has helped us break data out of the sole domain of analysts and data scientists and directly into the hands of planners, accounts and creatives,” Beacraft said.
Epsilon started its AI practice four years ago, and Beacraft said it has evolved “quite a bit” since then. That change has mostly involved building more robust means of data collection, honing proprietary analytics tools and formalizing processes.
“By turning the entire open web into the world’s largest focus group and leveraging our incredibly vast data resources,” he said, “we’re able to understand a brand’s true competitive set, build incredibly granular consumer profiles, expand a product’s use cases and jobs to be done and even reverse-engineer the strategy of successful new entrants to a category.”