The scene: Sitting on a large wicker sofa 30 feet from the Mediterranean Sea drinking Perrier in the Riviera sunshine and talking to the chief marketing officer of a bank about generative AI, synthetic audiences and creative optimization.
That’s how one rolls in Cannes.
But rolling back the clock to 2023, which is when Michael Lacorazza joined US Bank as CMO, the first thing on his mind at the time was to consider “the big strategic questions.”
“We wanted to rethink our approach to brand storytelling and our go-to-market approach,” Lacorazza told AdExchanger. “What could we do to create work that would more deeply and emotionally resonate with audiences?”
The traditional way to answer those questions would be to do standard market research and run a bunch of focus groups, pause, collate the data and then weave whatever insights were gathered into creative production and media planning – a process that could take six months or more.
US Bank wanted to move faster, Lacorazza said, and also bring in “some fresh thinking.”
And so it partnered with a startup called Supernatural AI that can create AI avatars of key target customers and shave months off the strategic development cycle.
“We’re able to get to the answers we need way more quickly than we could through a typical human process only,” Lacorazza said. “We’ve really learned a lot.”
Lacorazza spoke with AdExchanger, between sips of sparkling water, about US Bank’s ongoing investment in AI.
AdExchanger: How does the synthetic audiences thing work?
MICHAEL LACORAZZA: Supernatural created five different audiences profiles for us based on what we were going after demographically and psychographically. They used multiple third-party data sources to assemble the audiences and then trained them using different models so we could activate against them.
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
Activate against them how?
The simplest metaphor is a focus group. You can ask these audiences the same sort of questions as you’d ask humans and get responses back, but you’re doing it through a large-language model chat engine.
What sort of questions are you asking?
Say we have a synthetic audience of young, affluent, college-educated people in the first one-third of their career. Maybe they’ve also formed a household and they’re financially stable. We could ask them questions about what’s important to them in their banking relationship, what type of products they’re looking for and what sort of financial decisions they need help with.
But we can also ask questions about things that might be too sensitive to ask a human, like digging into the nature of how couples divide their finances. We can delve into tricky areas and get deeply personal without making anyone feel uncomfortable.
Do you trust the outcomes, though?
It was very important to us at first to see if we could validate what we were being told because, you know, these insights are being generated by a machine. So we did some tests with humans and the themes came back with a 90% to 95% overlap with the synthetic audiences.
Going forward, we’ll continue to validate, especially as models change and evolve, but we have a lot of confidence in what we’re doing. And, at the end of the day, humans are making the final decisions on all of the work we do. We always apply our own judgment.
How often are you getting pitched by vendors shilling new and supposedly “game-changing” AI tech?
Oh, every day. I mean, there are just so many startups and ideas in this space. It’s incredibly vast.
How do you decide who to work with?
Because of the scale at which we operate, we generally need enterprise-grade technology partners, and it can take a lot of time to go through a procurement process.
Meanwhile, the space is changing so quickly, so we’re not trying to chase a bunch of startups around so much as we just want to learn about what’s happening and be involved.
What do you think of Mark Zuckerberg’s plan to fully automate ad creation and personalized targeting using AI by the end of next year?
It’s a bold statement and I’m interested to learn more about how it’ll work. Meta has a vast amount of data on performance outcomes to support that vision.
But I’m not sure I’m willing to hand the keys to the kingdom over to a media partner to manage everything soup to nuts.
We think AI is good for audience refinement, generating insights and automation. The personalization use case is a little overhyped.
Is there an application of AI you’ve come across at Cannes that caught your attention?
I did a roundtable with someone who leads analytics and AI for a retailer and I was super impressed with what she’s been able to do in terms of automating processes inside her company. She said she’s automated something like 60 processes so far from end to end.
That really struck me. It might not be the sexiest thing, but it’s really impactful.
This interview has been lightly edited and condensed.
