Picture this: In the fast-paced world of media buying, artificial intelligence promises to streamline operations and unlock new efficiencies, but it also lurks like a double-edged sword, ready to complicate transactions and erode trust. This tension is at the heart of discussions among industry insiders—and it's sparking fierce debates that could reshape the future of advertising. If you're curious about how agencies are navigating this AI revolution, especially in programmatic advertising, stick around. We're about to unpack the latest insights from a candid Town Hall at Digiday's Programmatic Marketing Summit in New Orleans, drawing from anonymous chats under Chatham House rules that let attendees speak freely without fear of repercussions.
But here's where it gets controversial: Is AI a game-changer or just a fancy smokescreen for hidden agendas? As we delve into the nuances, you'll see why some execs view it as an unstoppable force for good, while others warn of its pitfalls. Let's break it down step by step, clarifying the jargon so even newcomers to the ad tech space can follow along.
At its core, this briefing explores the delicate balance between potential and peril when applying generative AI to media agencies. The summit, wrapping up Digiday's 2025 event lineup, revealed a landscape fraught with uncertainty. One key theme emerged: the blurred lines where AI intersects with programmatic advertising, particularly through agentic technologies—those autonomous systems designed to act on behalf of humans. Think of agentic AI as smart assistants that can make decisions independently, but participants couldn't even agree on a universal definition or how deeply it's already embedded in daily workflows.
What everyone did concur on, however, was a cautious approach: Keep AI agents out of the critical buyer-seller transactions. Instead, reserve them for supportive roles before and after deals, such as brainstorming campaign strategies or sorting through performance metrics post-launch. To make this work, agencies must invest heavily in training, not just on AI tools and their quirks (like their tendency to generate inaccurate 'hallucinations' based on flawed data), but also on the basics of programmatic processes. This way, human overseers can better guide and verify their AI partners, ensuring decisions aren't left to algorithms alone. For beginners, imagine programmatic advertising as a complex auction system where ads are bought and placed in real-time across websites and apps—AI could automate the prep work, but humans need to steer the ship to avoid costly mistakes.
And this is the part most people miss: The real challenge isn't just adopting AI; it's fostering a culture where employees deeply understand the 'why' behind the tools. Without that, we risk creating a generation of planners who can't think creatively or troubleshoot without relying on machines. But let's hear directly from the experts—I've compiled and edited their quotes for clarity and brevity, expanding on key points to illustrate real-world implications.
Setting the scene: One participant noted that while machine learning AI has long been woven into buying processes—think optimizations on platforms like Meta or Google—the real buzz surrounds generative AI and its agentic evolution. Agents that can ideate and act autonomously? That's the future, exciting yet undefined. We've already seen gains in tools that refine media mix models for planning, cutting down on manual grunt work in data aggregation and analytics. It's like having a tireless assistant that handles the boring bits, freeing up creative minds for strategy.
AI as a smokescreen: Every agency uses machine learning in algorithms and tweaks, but it's inherently opaque—hard to peek behind the curtain. Now, the industry fears AI is being hawked as a cover for even murkier pricing and performance claims. Take platforms like The Trade Desk's Kokai, which touts algorithms that 'do it all' and urge buyers to 'trust us.' This opacity terrifies pros who lack the tools to challenge it. Plus, AI doesn't explain its moves like a human optimizer can: 'Why this adjustment? What was the goal?' Without transparency, accountability vanishes. One exec shared ongoing efforts to build AI that demystifies campaigns, making the 'black box' more like an open book.
Handle with care, agent: Agents excel at data fetching and assembly, but they don't reason—they operate on probabilities, like gambling in Vegas for better outcomes based on past data. Don't expect them to innovate on novel tasks; they're not there yet. We're likely 3-4 years from true trust. For defined tasks, AI shines, but in programmatic planning, it might pull in outdated web info, leading to errors. Clean, reliable datasets are key—think of it as feeding your AI high-quality ingredients instead of junk food. With good data, it can evaluate parameters effectively; otherwise, results could be disastrous.
What's next on the horizon? From a strategy angle, some partners offer tools to craft entire media plans, but integration lags. The dream? One-button planning and optimization. Not yet, but coming soon—perhaps. Yet, enthusiasm clashes with reality: Legal risks, finance scrutiny, and the ever-present 'hallucination' flaw in large language models (LLMs) mean one wrong number could sink a client or career. LLMs are like brilliant interns—speedy on menial tasks, but prone to errors. Agencies often fall back on proven older models for critical work, valuing experience over flash. One analogy likened AI to a six-year-old eager to please, delivering answers regardless of accuracy—endearing but unreliable. For now, AI suits pre- and post-campaign phases: Planning teams handle ideation, analytics teams report outcomes. This saves heaps of time, as 'production' (the actual campaign run) is just a sliver of the workload—most effort goes into setup, monitoring, and tweaks.
Accountability matters: Managers must insist employees own their work, even if AI generated it—no 'the AI did it' excuses. Otherwise, errors multiply. A big worry? Younger planners (say, those in their 20s) might over-rely on AI, eroding deep knowledge for strategic, out-of-the-box thinking. Fast-forward 20 years, and we could have a workforce ill-equipped for innovation, dependent on tools without understanding the foundations.
Training your AI: Like slow-cooked barbecue, use mature, specific data for better results. It might be slower than real-time feeds, but cleaner—less 'garbage' means more reliable insights. Think of training as nurturing a garden: Patience yields stronger, truer fruits.
Shifting gears to broader industry trends—does the holiday frenzy create openings for retail media networks and upstart social platforms to snag more ad spend for 2025 and beyond? Absolutely, if they bridge the measurement and trust chasm, per fresh Kantar research. Key findings: Retailers prioritize in-store/omni-channel boosts (82%) and off-site data profits (45%) to rival Amazon. 87% of brands favor accredited networks with standard metrics (but only 24% of retailers meet that), 67% would pump in more cash if in-store effects are verifiable, and 80% deem unified online/offline measurement crucial. In simple terms, brands want proof that ads drive real-world sales, not just clicks—AI could help here by refining data, but only with transparency.
Quick industry updates: Omnicom's IPG deal fell through, sparking 4,000 layoffs and public outrage on social, plus the end of IPG's Magna unit and exit of top exec Eileen Kiernan. Havas scooped up UK experiential firm Bearded Kitten (into Havas Play) and French data outfit Unnest (for tech boosts), prices undisclosed. Personnel shake-ups: Dentsu appointed Kara Osborne Gladwell as global product architect for Media and reinstated Tia Castagno as Global Innovation President, both new roles under Will Swayne. Monks brought Thiago Correa as EMEA SVP of Media from Publicis (ex-H&M lead). Creator agency Influencer named Ryan Fitzpatrick CFO, from VaynerX.
A direct insight: 'This merger is a reminder that scale doesn’t fix fragmentation. AI only works when the underlying customer data is unified, governed, and understood. The organizations that win in the next phase of marketing won’t be the ones with the most tools—they’ll be the ones with the clearest picture of their customers.' —Tony Owens, CEO of Amperity.
Speed reads: Seb Joseph and Sam Bradley examined The Trade Desk easing agency fees amid DSP competition. Ronan Shields chatted with Criteo CEO Michael Komasinski on rebranding to AI-driven commerce via LLMs and M&A whispers. Krystal Scanlon forecasted a booming creator economy in 2026, packed with stats.
So, what's your take? Do you see AI as the savior of media buying, or a ticking time bomb of opacity and over-reliance? Should agencies prioritize human oversight to preserve creativity, or lean into automation for speed? Share your thoughts in the comments—does this spark agreement, disagreement, or fresh ideas? And here's a controversial twist: What if AI's 'hallucinations' aren't flaws, but opportunities for unexpected innovation? Let's discuss!
