Production, analytics, measurement: Meta and marketers mull AI’s use cases
Meta has explained its approach to AI tool development at the Meta Marketing Summit – here’s how the industry could harness it.
Meta believes AI is transforming performance, creative and distribution of content for brands
The night before the AI-focused session at the Meta Marketing Summit, Meta founder Mark Zuckerberg announced that the company had a dedicated division looking at generative AI. It is proof enough that the hype cycle around AI and marketing is well underway. Meanwhile, a panel of marketers and AI specialists was quietly dismantling that hype and focusing on the tangible benefits for marketers and brands.
Martin Harbech, group director at Meta, kicked off the session by noting that most technology breakthroughs happen slowly. For that reason, those moments can be so difficult to miss. He said: “It’s increasingly clear we are looking at one of those paradigm shifts. There’s a huge amount of hype; the important point here is that hype and impact are not mutually exclusive concepts.”
Angie French is director of marketing science at Meta. She explained that while there is a tremendous amount of hype at the moment, most big tech companies have been investing in AI as part of their ongoing services for many years. She believes that while marketers should be open to experiment, their strategies should be predicated on demonstrating point solutions delivered by AI.
She said: “We are really moving forward with how we think about machine learning and how that supports measurement in the industry, how we open source… It’s very much around point solutions… you have to be very careful about where you’re making those investments.”
Mark Zuckerberg says Meta now has a team building AI tools and "personas" https://t.co/9jsXyakdC5 pic.twitter.com/9n6O10qvOo
Mark Zuckerberg says Meta now has a team building AI tools and "personas" https://t.co/9jsXyakdC5 pic.twitter.com/9n6O10qvOo— The Verge (@verge) February 28, 2023
The panel was keen to point out that hype and impact are not mutually exclusive. Stuart Bowden is global chief strategy and product officer at Wavemaker. He said: “We’re at all stages of the hype curve simultaneously. It’s the speed of that feedback loop that makes it so different. The real impact for us comes in narrow areas… inter-creative production, creative analytics, all of those things that feel less exciting out of the industry. That transition to impact will be much faster than we’ve seen in previous hype cycles.”
The “less sexy” part of the discussion, per Meta’s global planning director Ian Edwards, ultimately ends up being the most transformational for the industry. He stated that since the company had rebuilt its tech stack to better integrate AI in 2020, it had seen a 20% increase in conversions in Q4 last year.
The panel noted that while once a planning document might have taken three interns three months to put together a single iteration, AI can effectively replace that process and iterate multiple times of a brief in a fraction of the time. Edwards cited analyst Benedict Evans as saying it was like having “a million interns” working on the brief.
Harbech notes that there is a joke in AI that “AI is everything that doesn’t work, because as soon as it works it becomes utility”. For French, that gap still exists when it comes to using AI for performance analysis at scale. She said: “What I’m interested in is performance on the back of that synthetic created media. We’re not quite there yet with integrating that into the wider marketing process. I’m looking forward to closing that flywheel… we’re really fuelling power and performance and we’re not there with scale.”
Dr. Alex Connock is a generative AI expert at the Said Business. He noted that we are at the foothills of what generative AI can offer to creative marketers: “If it was about distribution I would think about… how we think about creating content assets. Editing you now have systems where you can just say ‘show me the bit where she doesn’t like the car’, saving you three hours of scrolling through the rushes. That ecommerce revenue of six pieces of content a day will be… the smart advertisers will be on top of that.”
One example of creative use of AI cited by the panel was that of synthetic humans being embedded into marketing materials. Bowden cited research that demonstrated that, even when audiences were told they had been served up an ad featuring a synthetic human, most of the audience actually preferred that artificiality. He said: “Even when we know it’s fake, we trust people who look like us.
“It breaks down the relationship between publishers and content creators… suddenly cracks appear in the fundamental way we understand how the web is monetised.”
For Edwards, one of the biggest opportunities for marketers and brands lies in the evolution of relationships between AI and the consumer. He stated: “I think bots have been a bad experience up until now for customers. This next generation of conversational AI will transform customer relationship marketing. We’re starting to see ChatGPT take all the terms and conditions and [answer consumer questions] in a dynamic and human way.”
He also pointed out that consumers’ use of chatbots allows marketers access to zero party data - data offered up by the consumer themselves directly.
The panel did note, both during the main discussion and in the Q&A section, that while all those things are possible, they are not necessarily ethical or desirable. They noted that it is beholden on the heads of companies to both upskill themselves in the tech and to consider what impact it will have on younger entrants looking to enter the space.
The gap between hype and effectiveness around AI for marketers is still vast, but the lion’s share of that hype lies around tools like text-to-video, which the panel promised would be able to deliver a new season of Game of Thrones from a single sentence prompt within a few years. But the bigger revolution for marketers is happening quietly, behind the scenes, in terms of automation, iteration, and measurement.