Overheard at the Digiday AI Marketing Strategies event
Marketers, brands, and tech companies gathered at Digiday Media’s AI Marketing Strategies in New York City on Wednesday to discuss AI usage and ethics. As more and more companies incorporate various tools and chatbots into their daily workflows, we wanted to see how brands and marketers were leveraging these tools, and how their employees and customers feel about them.
The biggest topics? Training workers to better utilize AI rather than wholly rely on it, how publishers can use agentic tools, the importance of human touch and how to bridge internal divides around AI.
The conversation was conducted under Chatham House rules so Digiday could share what was said while maintaining speaker and attendee anonymity.
This has been edited and condensed for clarity.
What about the AI bubble?
“If you look at the dot com bubble, that’s the best example that we have of what happened in the past. A lot of companies went bankrupt, but there were some companies that actually did really well, like Amazon, for example. So I think [that’s] what will happen with AI. There are use cases that are tangible and real, and there’s a lot of hype… post-bubble, that hype is not going to last, but tangible results that actually solve problems and then lead to results that team will remain and will thrive afterwards.”
How companies are navigating internal AI tension
“With younger people on our team, they use the AI tools and it spits out a response or an answer. They’re junior and they don’t know how to analyze if it’s right or read between the lines or critique it — that’s probably been our biggest challenge.”
“Any junior person who does a deck using AI or you can tell something is crafted with it, it’s always reviewed with a more senior member of the team, and they walk them through and ask ‘Hey, where’d you get this information?’ It’s drilling in: AI isn’t always the best place to go for things, and kind of having that mentorship with the younger generation, just making sure they are reading, rereading, and then what to look for when using AI. Education is the biggest part of it.”
“We embrace it. We want everybody at our company to start using these tools… if you don’t start experimenting, you’re going to get left behind. It’s just doing so in a way that aligns with our clients’ brands, that doesn’t come off as fake and actually is giving correct information. So we do want them to experiment, but we also want to educate them on what could go wrong and what to look out for.”
“Even at a high capacity of understanding AI, there’s a misconception about it. That made me think today, I need to teach my team how to fact check and be more of a skeptic… So the first thing is, the common basics of education, where does this scrub from? Show them how it scrubs, show them the sites it scrubs from, show the long form it comes from, and then have them help read citation… I think all of us as leaders just go back to the basics that we all learned when we were in the library, and how we learned about our papers and things we wrote, and then educate our teams. And that’s whether they’re 65 years old, or they’re 23 years old.”
“We’ve talked about people who are very afraid of AI hallucinations. You’re going to get them. AI is like an intern, but it’s an intern that can actually help you look great, or you can act old school and kind of come down on that intern that hallucinates and doesn’t give you what you want.”
Using nontraditional platforms and creators to build brand awareness
“I’ve been in the industry 30 years and we use LinkedIn as a way to network… I think the biggest game changer in the two years for our company and our trajectory, our growth — we had a phenomenal two years — is our LinkedIn. We really are controlling the narrative, and I think the industry is really paying attention to it. We found some great vendors from it. If you’re a brand and you have an opportunity, and obviously if you’re a lean team, I know it’s like one more thing, but don’t shy away from LinkedIn.”
“One thing we’re also looking at is like the nano influencer as a source of review. So instead of always looking for that big like Kardashian post, like the trust people have in the more nano voices as more of that organic review and even opportunities where we can see the product and then they can provide that review… the trusted reviews are through gifted and paid opportunities.”
Different ways to maximize GEO
“From a retail search standpoint, or just retail PDFs in general, what we’re doing is we’re refining our product detail page copy to feed into the engines. The brand story is the brand story and the product story is the product story, but we are bringing in keywords that are very friendly for the bots to pick up on and then leverage in results. So that’s been a big work stream for us.”
“Awards are also a big deal in our space… We have specific comms where we create short clips with the award badge opening or closing, because for us, we know that our core audience is turning to those trusted partners for recommendations and for guidance… So it’s definitely an integral part in our comms… and we’re seeing that come through from an organic perspective.”
How publishers can leverage AI
“The publisher that you represent, if you’re using programmatic pipes to fill some of that demand, you set a floor price. You say ‘my inventory is worth no less than $10’ (making up numbers, of course). And so everyone who’s bidding on your inventory must bid $10 or above to win the impression. But not all impressions are equal, right? There’s going to be impressions on your site, no disrespect, that are worth 30 cents. But then there’s also situations where, if someone was shopping for Bounty on Instacart, they didn’t buy it, but in that moment, you know, they want to buy paper towels in a very short period of time, and you have that person on your site, Bounty might be a $30 CPM for that impression. Publishers have historically set a floor and a ceiling that are pretty close to each other, and in some ways, you’re underserving what you can make, and in other scenarios, you’re limiting the ability for lower value impressions to be sold, because on the buy side, they’re like, it’s not worth it.”
“We actually want to help publishers fix this, make it more just like ticket prices. Dynamic pricing is how I want you to start thinking about your inventory. You’ll then understand which impressions are worth more, which will actually trigger how you handle your site, your stream, or whatever it is you represent, to try to create more high-value impressions versus just impressions.”
“As a publisher, you sit on a tremendous amount of valuable data, but many publishers are still holding on to that data, because the notion of providing open access to that data feels like it might be shipping what is left of your ability to sort of retain autonomy and your existence to a level. But I think we as an agentic broker are attempting to get publishers comfortable is the idea that if you integrate your data with our agency data systems, it creates a more perfect connection to appreciate the value of your audiences and the value of your inventory.”
Are there strategic limitations to AI?
“Agentic can understand reams of data quicker than any human can once you train it once.”
“Every brand in the world wants to maximize their working media dollar. 99 percent of the time a brand will get what they call log level data, which is basically showing where all the money is going. But you have to understand how to analyze it, and you get it at the end of the campaign. Twenty years ago, you’re great. But five months ago, right? It’s too late. It’s way too late. Agentic can go through that on a daily or weekly basis and give you better summaries.”
“Where we are going, from an agentic standpoint, is that we now live in a place where segmentation and re-segmentation and segmentation powered by in-the-moment modeling intelligence is able to create new audience packages that get sent and refreshed to our targeted line items in real time, such that, the return on the quality of those audiences (assuming that your predictive modeling is actually strong) ends up being sort of significantly better, because no longer are these static segments, they are dynamically refreshed, pushed smart audience compositions, and at scale.”
