Generative AI is the talk of the town. At creative and media agencies, everyone from designers to copywriters to marketing leaders are giving generative AI programs like ChatGPT and DALL-E a try.
The latest iterations of “incredibly user-friendly” generative AI tools “have set things on fire,” said Emma Cookson, a partner at mar tech holding company The Brandtech Group.
Due to generative AI’s startling capabilities – such as writing marketing copy and generating beautiful (or bizarre) images based on instructions – it’s encroaching upon the creative work people previously assumed only a human could do.
“It genuinely looks like one of the most disruptive and enduring technologies to impact this business since the mobile phone,” Cookson said, a reflection of the complicated, multiyear developmental work that has gone into building and refining AI models.
Buzzy though it is, generative AI is far from reaching the singularity, or the hypothetical point when superintelligent AI improves itself with such blinding rapidity that humans surrender to their robot overlords (and eventually, perhaps, go extinct).
For now, generative AI tools are shiny new toys that many agencies can’t resist unboxing and playing with, judging from conversations with more than a dozen industry professionals. But opening a Pandora’s box of AI toys could release real risks into the agency world, such as AI upending the traditional full-time equivalent (FTE) agency payment model and running afoul of intellectual property and copyright laws.
Still, “these technologies are fun to experiment with,” said Ellen Jantsch, CEO of growth marketing agency Tuff. “It’s worth trying them out to see what they’re capable of.”
Helper bot
Content marketing can be “grueling,” according to Jantsch. Researching topics and creating a list that feels on brand and uniquely relevant to readers can take hours. ChatGPT can generate 100 topics for creative strategists to use as a jumping off point before narrowing things down to 10-15 topics.
The creative team has tested ChatGPT’s ability to generate video marketing scripts, while the media team has used the program for keyword research.
“We’ve been able to integrate the tech into existing processes and make the output a little bit stronger,” Jantsch said. “Certainly not a replacement, but a good value driver.”
ChatGPT struggles with developing a “branded tone of voice” and can be “quite robotic,” said Simon Spyer, CEO of data-driven marketing at digital performance marketing agency Iris Worldwide. But it excels at SEO.
“[Search is] about having the right tags in place, the right keywords, not being too repetitive in your copy,” Spyer said. “ChatGPT works really well in that environment because it’s referencing a piece of content to that rule set and helping generate something [that matches those criteria].”
On the image-generation side, said Ian Schatzberg, CEO and founder of brand design agency General Idea, the agency has been using DALL-E for visual research and exploration.
For a 3D billboard project, for example, the team put the subject matter and an object into the AI, which generated hundreds of pictures in response. The team then edited the pictures and used them to conceptualize the billboard build.
“[AI] illuminates new ways of seeing things that are familiar to us,” Schatzberg said.
Sometimes too new. When the AI spits out images “divorced from reality” that don’t make any sense, Schatzberg said, a person has to step in to “humanize” the outputs.
And not just any person. Training the AI “requires a really deep visual vocabulary,” Schatzberg said. “What you feed it is what it trains on. If you have a broad visual diet and extensive visual references, you’re better equipped to train the tool.”
Rewriting the agency
With the rise of AI, is an agency shake-up nigh?
AI will disrupt the advertising industry more than the internet did, more than TV did when it shook up a radio-dominated advertising landscape, said Mark Singer, US CMO at Deloitte Digital.
Yet he’s confident agencies will adapt. “Agencies are not going to be replaced by robots,” he said. “But AI is going to be a big part of the future of the agency.”
AI could remake agencies the way Photoshop altered the once specialized, profitable retouching trade, which had inaccessible tools and a high barrier to entry, General Idea’s Schatzberg said.
Today, the retouching business is much more democratized. It also has smaller margins. Comparably, AI could make the creative process more accessible, thus lowering both the cost to entry and the margins on certain services.
“It will cause a sharpening of the role of the agency as arbiter, curator and conduit into culture,” Schatzberg said. Agencies will be responsible for the intentionality and ideas behind the work. AI will handle the creative labor.
Jon Bond, founder of marketing consortium Weightless, predicts agencies paid through an FTE [full-time equivalent] model, which calculates agency compensation based on employees’ billable working hours, will face problems in the near future. “Agencies aren’t paid for the big idea,” he said. “They’re paid to execute the idea.”
If an agency is paid $1 million – $100,000 for an idea, $900,000 to “make 87 iterations” of the idea – and those 87 iterations are generated by AI, it gets much harder to justify the hours or people on the project.
“I’m glad I’m not a junior copywriter starting in this business,” Bond said.
In the next couple of years, “the hottest job is going to be AI trainer,” Bond added – someone who partners with the AI, giving it specific prompts, guiding it and teaching it to get better at shared tasks and projects. The new model for a creative team will be a human and a machine, collaborating as closely as an art director and copywriter.
Credit where credit’s due
But when a generative AI tool and its trainer are so intimately intertwined, who gets credit for the work?
Traditionally, advertising doesn’t come with a credit list, said Ben James, chief innovation officer at digital marketing agency Gale. No credits roll at the end of a 30-second commercial, and print ads don’t name the photographers.
But the brilliance of generative AI often comes from ripping off human creativity. Generative AI is raising many creators’ hackles, and lawsuits from those artists are starting to accumulate. For example, Getty Images is suing Stable AI, the maker of Stable Diffusion, for copyright infringement, and three artists are leveling the same charge at Stability AI and Midjourney.
And IP violations are hardly the only concerns occasioned by generative AI.
“AI has no concept of truth,” said Aaron Shapiro, founder and chairman of indie agency Product.
Lacking a built-in mental model to discern between truth and fiction, AI can glibly produce authoritative, interesting, completely inaccurate copy – and have no idea it’s wrong.
“It’s a long road between what exists now, which is a nonthinking tool, and something that adds the value that humans have,” Shapiro said.
