Generative AI Presents Small Agencies an Opportunity to Scale

As conversations about the Metaverse and Web3 were swirling, Aaron Shapiro, having recently stepped down from his role as CEO of the IPG agency Huge, noticed another technology quietly gaining traction in the creative community. Agencies were paying more attention to generative AI, or auto-generated text and imagery. For Shapiro, this felt different than when AI dominated the conversation years ago, amid launches of voice assistants Alexa and Siri. “AI was very much a novelty then,” he said. “When you really peeled back the onion, AI wasn't really used for any concrete, real things.” But generative AI was different. Its relative ease-of-use tore down boundaries that make emerging technology inaccessible to all save those with the largest budgets. By contrast, OpenAI this week announced its premium subscription version, which costs just $20 each month. The technology had disruptive potential, even though as far as Shapiro could tell, the way most agencies were working hadn’t changed at all. So last year, when he founded his new marketing agency, called Product, Shapiro embedded generative AI into the agency’s workflow. Small and mid-sized agencies are beginning to use the technology to produce more creative work faster and with fewer hiccups. Meanwhile, clients, big and small, are asking agencies of all sizes questions, and media technologists wax poetic about the technology’s potential to upset workflows. The industry is entering a “creative renaissance,” Mike Liu, head of innovation at Dentsu-owned agency Carat, told Adweek. How Brands and Agencies Are Experimenting With ChatGPT From Copywriting to Chatbots A ‘test and learn’ strategy for everyone Product uses the technology to test creative imagery at scale. A so-called test and learn strategy can save clients money and increase their ROI. A lot goes into creating assets like this that only flop upon activation. Assume that a brand invests hundreds of thousands of dollars creating an arsenal of ads with blue backgrounds, only to find that once they’ve activated the media, they find that the target audience just doesn’t like the color blue. For most agencies, practical limitations leave little room for agencies to truly test, learn and alter their campaigns in real time. Agencies often work in silos, and this combined with low bandwidth, makes it tough to match the right creative assets with the right audience and optimize that creative depending on the audience’s engagement, or lack thereof. But AI-generated work has the potential to change this, according to Shapiro, assuming that agencies use AI to produce low-fidelity versions of an ad and activate that creative on social platforms to small subsets of a large target audience. “You could test right away to see if messages work,” Shapiro said. “And not only that, you could do this for dozens and dozens of derivatives of concepts—testing color, style, images, types of text, tonality, whether it should be a live image or an illustration,” he added. One large consumer company tapped the agency to help it launch a product. Before the product ever launched, the agency used AI to auto-generate a catalog of messages and evaluate their effectiveness. After testing the messaging with a small audience, it discovered what resonated most with the brand’s target audiences. “That allowed us to do a full launch with pre-vetted messaging,” Shapiro said. A "test and learn" media strategy is not a novelty on its own. For example, VaynerMedia executes this strategy, which it calls its Vayner Volume Model. But executing a similar strategy is daunting, if not impossible for many agencies that don’t have a similarly deep expertise in social media or an integrated creative team on hand to mass-produce different variations of creative assets. Now, testing out creative concepts en masse is one of Product’s core service offerings. Earlier this month it launched its AI Studio, overseen by Tim Nolan, the agency’s product partner and executive creative director. In a statement, Nolan said he refers to tools like Stable Diffusion, Midjourney and ChatGPT as his “junior team.” In fact, every project the agency works on is supported by AI to some degree, its leader told Adweek. “It’s just part of the process in which we produce work,” he said. When used to support a campaign strategy, the technology can cut down the hours creatives spend sketching a concept or hunting down stock photography that illustrates it. Now, they can use generative AI to almost instantaneously produce images that represent their ideas. The opportunity looks different for holding companies Liu, whose role at Carat revolves around educating clients on the latest technology trends, described AI as “ramping back up again,” following ChatGPT’s mainstream availability. About 75% of Dentsu Media have approached Liu with questions about the technology. He plans to spend the next three to six months educating those clients ahead of those brands taking action. Conversations taking place at Liu’s level revolve around to what extent AI-generated creative will be consumer facing versus a behind-the-scenes boon to campaign strategy. “When we pivot into this big crescendo of generative AI, from text-to-image and image-to-image based AI, a lot of our clients are [wondering], 'Can this change our workflow of building creative?’” Liu said.But Liu's clients are only beginning to understand what the technology can offer, he added. Meanwhile, Dentsu is building its knowledge, with media and creative work streams converging. Liu is having frequent conversations with his innovation counterpart at Dentsu Creative, foreshadowing how large networks must rally to establish generative AI strategies that reach all sides of the house. Regardless of how agencies use the technology right now, its inherent flexibility makes it a worthwhile addition to creative workflows. “Those [agencies] who have specialists on their teams that understand how to utilize these platforms as this toolbox, or something that can help augment your workflow, will greatly have an advantage over those who don't,” Liu said.