I’ve had this Victor Hugo quote stuck in my head these past few days, “nothing is more powerful than an idea whose time has come.” It’s great writing of course, but It feels true here, in this nondescript Hilton reception bar area where I write this. Word is we’re going back to the Marriott next year thank god. But I digress: 2026 is the year AI irresistibly takes over advertising.
So here’s what you missed if you didn’t make it out to RampUp, LiveRamp’s premiere homecoming ball for past and future LiveRamp employees (six and counting at PurpleLab alone): AI. AI AI AI AI AI. It actually started last week I’d say, with SemanticIQ’s hard launch – read their launch announcement, it’s good.
I know what you’re thinking: Ted, we’re already talking so much about AI. You burst in through the wall like the proverbial Kool-Aid™ man, as we’re all enjoying a rich supper of AI, to try to get us to drink from yet more content about AI from you jug? Didn’t you just publish a piece on the NAI about AI a week ago? Yes dear reader, I did, but somehow you still can talk more about AI.
As it’s been said many times, adtech’s run on AI (machine learning, specifically) for years, what’s changing now is how it’s being positioned to change the workflows of those working in the advertising field, rather than just being a slightly different engine under the hood. AI is no longer behind the GUI, it’s here in your face and it’s visibly excited to reduce your headcount by 20%. Or to supercharge your growth so maybe you don’t have to lay people off. That’s one thread I kept seeing throughout- much is so new that there isn’t a settled tone, a leitmotif that indicators this new presence is a blessing or the great devourer of jobs. Either way it’s here, and there’s hardly any time to get used to it.
LiveRamp’s CEO Scott Howe was in rare form for the keynote, which began with what I might politely term fundamentals- “insights drive progress, more scale=more insights”, but dove into the new directions LiveRamp is pursuing with interoperability, presumably powered in part by it’s Habu acquisition and new head of product Matt Karasick. This is where things picked up steam, where LiveRamp makes it’s bold new play with AI: it’s no longer just the place where you go to send data somewhere else, with agentic tools, it’s becoming the place where you’ll build things, analyze performance, collaborate, report. If I were a LiveRamp shareholder I’d be heartened.
So what tools are we using? What companies are leading the pack? Here’s where the wave of AI really hit- it's the boomtime for startups purpose built to capitalize on the agent wave. The three widest heard names at the conference all have pretty different applications
SemanticIQ, from the team that brought you PlaceIQ, is building the interpretive agent for orchestrating datasets. Chalice Custom Algorithms, from power couple Ali Manning and Adam Heimlich, are offering the ai powered container for supply side media optimization. Newton Research, from John Hoctor founder of LiveRamp acquisition Data+Math, is building nature language tools for measurement.
I love when founders iterate with new companies as technology evolves- it's how we got PurpleLab. While these three took literal center stage, numerous other players (Akkio, Integral, PurePlay, AboveData to name a few) flocked the halls in a way I can't say I've seen for a few years.
I’ve been described by a few as needlessly cynical, I prefer "seasoned", but typically this stage of the Gartner hype cycle is labeled "the peak of inflated expectations", as contrasted by the following stage I live in, "the trough of disillusionment". It's comfortable in the trough, low expectations are easier to deliver on, but Anthropic’s Opus 4.6 has broken the back of my skepticism. Testing the latest models I fear my beloved trough may not emerge, and that this technology may fully deliver on it's promises.
That's certainly what a lot of the big boys think. Down the street from RampUp at the considerably nicer digs of the St Regis my and a great many other peoples CEOs gathered for Travis May's AI in Healthcare Summit, now in it's second year and a harder ticket by far to get. Travis cofounded LiveRamp and founded Datavant before launching his family office/ venture fund Shaper Capital, and his event serves as a nexus for big ideas about how tech can shift our healthcare system. I've long observed the similarity in physician engagement to B2B marketing, where prescribers are proximate to CTOs in the eye of a brand, similarly here I was able to see the overlap between adtech and healthtech. The answer, as Andreessen would word it, is that agents are eating the world in both cases.
I just dropped in to collect some notes from the boss and sample a cocktail (Raj at the bar makes a fabulous Last Word if you get a chance) and got pulled upstairs to see Vinod Khosla interviewed about the shape of the world to come. Amid observations about precision medicine accelerating launch timeframes and driving costs down for sickle cell anemia, he tossed out that the level of performance of medically trained agents today is close to surpassing doctors. What's the right price to pay for an eight year residency if I've got a model that can diagnose the patient as well as me?
Certainly not a question that I, with my brief liberal arts pedigree, am equipped to answer, but the clear parallel to adtech looms- as agents become capable to the point of executing the ad buys and measuring the result, what's the line between enabling and replacing the human? Speakers at both conferences were quick to stress the need for human in the loop workflows, either to be able to integrate communication into the care journey empathetically or to safeguard against hallucinations. But when I think about a machine replacing a job, I'm reminded of the one about the two guys in the jungle who see a lion, and one start's lacing up his shoes. "Why bother, you can't outrun a lion"- "I don't need to, I only need to outrun you". If the only reason I'm keeping my job is because I stop the machine from messing up, then once that machine has an error rate lower than a human it's curtains for me.
This emerged again on one of the last panels I caught with the founders of Newton Research and SemanticIQ, where both were asked why their tools needed to be adopted. Their answers were two sides of the same coin: one, the threat that there will not be more resources, and that you must find ever more ways to squeeze more out of less; the second that with AI answering the basics questions so well, you’ll be able to leave no stone uncovered when thinking of new ideas, that the "one last thing" an overworked analyst never had time to do, that could yield the essential insight, will be within reach.
I can't help flipping this coin back and forth in my hand, hopeful for the sake of the many familiar, friendly faces I saw at the conference that the doom and gloom reason doesn't win the day, but knowing that fear moves sales. For my part though it's been a blast- after all, the AI needs data to do all these things, and I'm selling it like it's going out of style!
Originally published on LinkedIn.