Five things marketers should know about generative AI in advertising

The creative industry is slowly beginning to use the technology, captivated by its ability to spin out ideas and mop up drudge work, but leery of its sporadically kooky output.

A still from Mint Mobile’s recent ad, featuring Ryan Reynold’s reading a ChatGPT generated script.​

Many people have watched with delight and disquiet as generative artificial intelligence (AI) has reached new levels of sophistication. ChatGPT maker OpenAI last week released GPT4, the latest iteration of its tech, calling it the most capable yet. Marketers are intrigued, but cautious.

The ad industry has already experimented with AI to help with creative work, but the arrival of viral chatbot ChatGPT has marketers trying to figure out how they can effectively use the kind of generative AI that underpins it – systems that produce content from text to images.

Here are five things marketers should know about the technology.

1. Hatching ideas

Jones Road Beauty, a makeup brand founded by cosmetics pioneer Bobbi Brown, leaned on generative AI for a video shoot this week using a concept developed by the productivity and AI company Notion AI.

Cody Plofker, chief marketing officer at Jones Road, expects generative AI to be a valuable way to come up with ideas and work more quickly. The technology could cut the time it takes to produce the dozens of ads a week that Jones Road runs with Meta Platforms Inc., which owns Facebook and Instagram, Mr Plofker said.

“You can say, ‘Give me 10 ad angles for Miracle Balm based on customer reviews,’ or something like that,” he said.

AI may also help brands battle “creative fatigue,” said Wesley ter Haar, executive director at digital advertising and marketing services company S4 Capital PLC and co-founder of S4 company Media-Monks.

“One of the key benefits of generative AI is its ability to generate large volumes of content in seconds, which means you can visualise and test out different concepts at speed,” he said. “You can also easily swap out different elements – like changing the background, changing characters – at any point in the production process without having to go back to the drawing board.” The tool may help agencies and brands do more with fewer people, Mr ter Haar said.

“I think it alleviates some of that head count pressure,” he said, adding that he believes the human element will always be there and that AI doesn’t fully replace a human’s perspective or skill set.

2: Reducing busywork

With generative AI, ad creatives will spend less time on rote tasks and more time working strategically, marketing experts said.

“Some of those mundane tasks, some of those things that have traditionally been done by hand, so to speak, those go away,” said Paolo Yuvienco, chief technology officer at ad holding company Omnicom Group Inc.

Jones Road Beauty’s Mr. Plofker said generative AI could largely relieve his marketing team of things like analysing customer reviews.

“If you can have AI read all of our customer reviews, and then kind of consolidate that and give a summary of what people are saying, good and bad – I see tremendous value in that,” he said.

3: It’s not replacing creatives (yet)

Mint Mobile, the wireless carrier that has agreed to be acquired by T-Mobile US Inc., in January released an ad featuring Ryan Reynolds – a part owner – reading an oddly off-key script written by ChatGPT. The ad created buzz for the brand, but got its power more from the copy the AI served up than from its ability to produce creative messaging, Mint CMO Aron North said.

The ad got “an emotional response – it’s shock and amazement that a computer can do this,” Mr North said. “But I don’t necessarily feel other emotions, like love or connection or humour. I’m not laughing with it – I’m laughing at it.” The new capabilities have invigorated a long-running discussion about the potential for automation in advertising, but haven’t changed the conclusion, according to Jay Pattisall, a vice president and principal analyst at Forrester Research Inc. While some might look at generative AI’s ability to write prose, for instance, it is only generating output from data and not originating thought or expression, he said.

“There has always been a tension between the art and the science of marketing and creativity, even when the technology was much more rudimentary than that of generative AI,” Mr Pattisall said. “Machines in whatever form simply just can’t replace or replicate human creativity. That was true then, and it is absolutely true now.”

4: Stay vigilant

Marketers need to closely vet the data sources that train their generative AI models, and monitor the accuracy of their output, Mr Pattisall said, citing risks such as an advertisement with incorrect prices.

There are questions around ethics and bias that marketers must be aware of, Mr Pattisall said. If generative AI is trained on specious or biased information, its output will reflect that, possibly churning out offensive, incomplete or outright false content.

Omnicom’s Mr Yuvienco said that although it is leaning into and deeply exploring this area, the ad giant is cautious about using generative AI to create consumer-facing work given all the questions.

“When it starts to generate things that are visible to our clients, that are visible to consumers, that’s when you start getting into some tricky legal things, ownership, biases, all the things that you can’t necessarily control or don’t have a good handle on today,” he said. Among those questions is who owns an idea if it came from generative AI.

“If a machine creates it, how much do you need to actually alter the output from the machine in order to then take ownership of it? These things haven’t been answered,” Mr Yuvienco said. “It’s a little bit of a wait-and-see game.” There is also legal uncertainty around what generative AI learns from, and what happens when it is trained using material owned by others.

While human creativity also uses past work to build on new concepts, it isn’t clear yet how courts will treat the issue when it comes to generative AI, S4’s Mr ter Haar added. “I’m sure the lawsuits will be extremely interesting.” And the technology itself still has room for improvement. Insofar as it might be used to create imagery or ads, Mr Plofker says it’s just not ready for prime time.

“I’ve tested them. None of them that I have seen are there yet,” he said.

5: Carve out a place in the budget and organisation

Industry players are reorganising their teams to keep up with generative AI’s quickly changing capabilities.

Omnicom has created a “center of excellence” around automation across its agencies, along with an AI steering committee.

It uses these groups to explore what’s possible with areas such as generative AI and how it can affect its work, along with the constraints and legal implications.

Mint Mobile dedicates 10 per cent of its marketing budget to “unknown, unproven and untested” causes, according to Mr. North.

It is using that budget to experiment with early-stage technologies, such as generative AI.

“Because of that, we can assume risk that won’t hurt our business,” he said.

“When you have money set aside, and you also have teams that have the mind-set that is not risk-averse, but risk-embrace … they’re willing to go out and try things.”

The Wall Street Journal