‘There’s a lack of trust’: Agency execs discuss AI, cookie burnout and inventory issues during the Digiday Programmatic Marketing Summit

At the Digiday Programmatic Marketing Summit last week in Palm Springs, California, agency executives participated in closed-door town hall discussions, in which they were granted anonymity for their candor. And they were quite candid about the challenges they are having to confront. Here’s a sample of what was said.

AI issues

“I think if you asked 90% of agencies what AI is, they couldn’t tell you a straight answer. Each company has their own meaning.”

“If you just sit and say AI is gonna happen to you, the inputs will be bad, the outputs will be shittier, and fraud is going to take over.”

“We have a full-time job running our campaigns, running our programs and partnering with clients. To then also solve for AI, simply because we’re the dynamic nerds who can talk about it is not a good enough reason for us to take it on.”

“This is an R&D issue. Agencies are primarily funded for performance and procurement. So you have a fundamental problem in funding, actually training these models.”

“We’re doing contract renewals with some clients, and some are flagging, ‘Hey, anything that you’re using AI for needs to be explicitly stated.’ And there’s now caveats like, ‘Hey, creative, you can’t use AI here.’ Because there’s a level of, people are afraid. Inclusivity is not built into AI.”

“We’re not putting any financial data or anything that’s compromising the client’s identity [into the AI tools].”

“There’s still always that human element of — it’s not like I’m just gonna take the output and be like, ‘Oh, don’t have to think about this at all.’ To me it’s that time savings, but it doesn’t replace me. I still have to do my job to ensure accuracy, quality and all that sort of stuff.”

“The data privacy angle is why we don’t have [an AI chat instance] open right now, like we’ll probably have in a couple months. It’s because we’re not using ChatGPT because the data they have available. Our intellectual property, we get audited. ChatGPT, we can use for personal, but it’s no longer allowed across our org. We leverage [Google’s] Gemini because they have the rights and the privacy, so we’re able to maintain it as part of our Google Drive relationship.”

“From a conservative company that gets audited, they’re very, very hesitant to put anything like cost information into an AI unless they know it’s a closed bubble where competitors can’t get or nefarious bad actors can’t get.”

AI efficiencies

“Anytime my team is copy-and-pasting from one spot into another, that’s a last-mile solution that AI and automation need to help fill so [that] my team is spending less time trying to get the work done and more time trying to make the work better.”

“I’m happy to take on that AI development and thought-starter ideas for my team, because the efficiencies are how we get more competitive pricing.”

“A great instance of an AI project that I worked on was we just created a chat instance of all aspects [for campaign creative sizes]. Because every time you start a campaign, the creative team’s like, ‘What size banners do you need?’ And they never change.”

“There’s this software called CorralData. I can say, ‘What’s the top performing store out of these 200 franchised stores?’ And it will aggregate all the data [and] within like five seconds will start spitting answers out. That saves me so much time.”

“We work with a casino, and the creative team built [an AI chat instance] that helps them name casino promotions based on everything they’ve ever named a promotion, words that they will never use. And it doesn’t replace the creative hours, but it helps them at least refine versus going to the clients claiming, ‘Oh, we already did that four years ago.’”

“We work with Seedtag, and we’ve been using their generative AI creative to build out additional creatives for some of our brands that don’t have as many of their own. And of the campaign we were testing it on, it was the only that we had that was driving lift and consideration and awareness.”

“You can also use [AI] in identity reconciliation. If you’ve got a Ramp ID, a UID, a Google identifier and a [Conversion API] integration on Meta all trying to claim the same individual, you can have [the AI] train off that to create associations to help you refine whether you’re reaching someone in a fragmented place or you’re bringing them onto another area, to help with your reach approach.”

Cookie burnout

“Obviously you’re still working towards that deadline that cookies are going to be deprecated, but I think people are over it, to be honest.”

“Everyone talks about it, but no one’s thinking of the actual campaign implications of what’s going to happen because it’s been so long [since companies originally started preparing for third-party cookie deprecation].”

Measurement woes

“When it comes to targeting, we’re in a good place. A lot of agencies and brands are in a good place there. Where we’re struggling is measurement and attribution.”

“A lot of our dollars are shifting to the Metas and Googles of the world because they can say, ‘I put in $1 million and got $6 million out.’ Man, that’s a great story my finance team understands. But that ties our hands together because so much of our dollars are going there because of the gap on the measurement side to be able to prove out from a clean, perfect line.”

“If I’m going to tell my CMO, ‘I hired the best agency, the best team, but I can’t tell you what I used to tell you [in measurement reports],’ that is a huge problem that our industry needs to find a way to resolve and work through together.”

“The client has to be aware of changes coming in and how they have to also evolve and invest.”

“When we first started, I felt like digital was a guide to performance; it wasn’t a true one-to-one. And it’s turned into a last-click [attribution model], but there’s so many different factors that initiate that outcome.”

“It was always known that there’s people data and media data and then the stuff that comes out of your campaign. And then [the agencies] started buying the data companies, and now it’s in one place. It doesn’t mean the same thing, though. It doesn’t mean you can put it together. It’s like trying to put together Magna-Tiles and Legos. We’re now stuck in a situation where we’ve got our Magna-Tiles and our Legos and we’re doing like this every day [gestures with hands as if trying to combine opposing forces with each hand]. And we know it’s not gonna work, but they pay us to do it.”

Inventory issues

“There’s a lack of trust. We’re reaching out to partners who lied to us a couple of years ago [about how they were managing the programmatic supply chain]. You’re going to talk to people and some may lie straight to your face or may not know the answer themselves. So it’s a meeting we have, but it’s a painful exercise.”

“There’s so much resold CTV inventory now. You might be buying from Magnite on a PMP, and I don’t want to say anything bad about Magnite, but don’t be putting trash CTV in the bidstream.”

“[The Google Video Partners controversy] was a huge scandal. Buyers who were buying YouTube were really buying Google Video Partners, and it’s running on God knows where. They say it’s not MFA, but we did a URL pull, and there were some interesting sites on there.”

“[Supply path optimization] should be an ongoing conversation.”

“In one of the instances, specifically Forbes, we’re talking to some of the brands and GARM and TAG about the fact that actually [Forbes] knew about it. And so were they classified as regular monetization? Therefore, if it was under their domain header, if it was under Forbes.com, which it was, is it actually MFA?”

“[SPO] is always evolving, and we’re always evolving. It’s not something that could be considered settled. It should be looked at every year, for every account, for every brand.”

“We’re just playing whack-a-mole with all of the fraud and cheating stuff.”

“I don’t like feeling like an idiot, like you did a thing that was wrong because we’re not watching it closely enough. It’s whack-a-mole.”

“We pay [the verification firms] to protect our media. And now we have to come up with another solution to protect the media again. Where do we draw the line? And what are we being sold into?”

“The onus is on us as a community to push back on [the verification firms] to be like, ‘Okay, so you created an AI solution that you say is going to help us address this stuff; how are you teaching it? How are you updating it?’”

Exec 1: “It’s incredibly manual [having to crawl through campaign reports to check URLs that carried ads]. But we have to.”

Exec 2: “That’s what we need AI for.”