Sweat beading on my forehead in a San Juan conference hall, Beet Retreat 2026 buzzing around me, Holly Yonosko steps up and drops this bomb: AI guardrails aren’t chains—they’re the only thing keeping ad tech from optimizing straight into irrelevance.
Look, I’ve covered this circus for two decades. Data? We’ve drowned in it since the dashboard dashboards first lit up executive egos back in 2010. Three-quarters of buy-side leaders now call traditional measurement a dud—eMarketer’s February 2026 report nails it. But Yonosko, Omnicom Media’s chief analytics officer, isn’t peddling more dashboards. She’s pushing AI guardrails right out the gate, the kind that lasso wild algorithms into serving actual business outcomes like revenue or client retention.
“We have access right now to the most data we’ve ever had access to,” she told the crowd. “But then where the issue comes in is the measurement and the actionability of that data.”
Why Does AI Need Leashes in Advertising?
Here’s the thing—AI agents sound sexy, zipping through optimizations faster than a caffeinated trader. But without guardrails? They’re dumb sprinters, sprinting toward the nearest shiny metric. Frequency caps. Click-through rates. Vanity stuff that pads spreadsheets but starves the bottom line.
Yonosko’s pitch: Anchor those bots upfront. Tell ‘em revenue matters more than impressions. Mix in efficiency checks—attention curves, whatever—but prioritize what keeps clients paying bills. Ignore that, and your AI turns into a rogue optimizer, juicing whatever’s easy to measure. Clients wake up to campaigns that look great on paper, tank in the real world.
A single sentence: Speed without direction? Disaster.
And this isn’t new. Remember the early programmatic days? Everyone chased CPMs like it was gold. Agencies pocketed fees; brands got burned. History rhymes—today’s AI hype feels like that dashboard fever dream, repackaged with neural nets. My unique bet: If Omnicom nails these guardrails (and they might, with their Acxiom toys), independents get crushed by walled gardens like Google’s PAIR or Amazon’s retail media machines. Who profits? The data lords, always.
“It’s really important first to anchor your AI, if you’re building an AI agent for your client, in those business outcomes,” she said. “Ensure you’re defining those up front, whether it’s revenue, whether it’s retention of your clients, whether it’s growing clients.”
The Guardrails Problem: Efficiency Trap Exposed
Faster sounds good. Always does. But Yonosko’s waving red flags: Unchecked AI chases isolated efficiencies, blind to the big picture. “You need to make sure it understands what are some of those efficiency metrics that it should be aware of as well as those business outcomes of revenue or frequency or some of our attention curves,” she adds. Context, people. Without it, bots go feral.
I’ve seen it. Campaigns that hammer one channel, ignoring the cross-platform dance. That’s where Omnicom flexes—Acxiom’s Real ID, their unified identifier. Links channels, IDs unique folks, feeds into clean rooms. Privacy-safe, sure, but mostly a moat for their stack.
“We have Real ID, which is a unified identifier,” Yonosko boasts. “We can connect cross channels to identify unique individuals, to then also create a clean room and do measurement within a secure closed loop environment so that we can provide real optimization metrics and measurement for our clients.”
IAB surveys back the hunger: More ad buyers eyeing cross-platform lifts. Not channel silos—holistic consumer outcomes.
But cynical me asks: Who’s paying? Omnicom clients get the pitch, agencies like them bill the implementation. Brands? Stuck hoping guardrails deliver, or they’re back to fragmented hell post-cookies.
Is Cross-Platform Measurement Finally Here—or Just Another Promise?
Fragmented identities? Ad land’s eternal curse. Real ID promises the glue. Clean rooms let you peek without PII leaks. Brands analyze behaviors safely. Sounds tidy.
Yet, dig deeper. IAB says buyers want it—planning more focus. Yonosko echoes: “We’re not really looking to understand how does each individual channel perform, but what is the outcome of having these channels work together on each consumer.”
One punchy truth: Channels alone? Meh. Synergy? That’s the money shot.
Still, skepticism reigns. We’ve heard unified ID gospel before—LiveRamp, The Trade Desk’s UID2. Many wilted. Omnicom’s edge? Owning Acxiom outright. Vertical integration, baby. But does it scale beyond their fortress? Or just lock clients in?
My bold prediction: By 2028, guardrail-first AI frameworks become table stakes, but only giants like Omnicom thrive. Indies scramble or sell out. Retail media eats the scraps—Amazon’s already there, measuring everything in-house.
This Beet Retreat moment cuts the PR gloss. Data abundance? Curse if mishandled. AI? Tool, not savior. Guardrails? Smart cynicism, forcing tech to serve humans (or at least their P&Ls).
But who really wins? Follow the data flows—straight to agency coffers.
🧬 Related Insights
- Read more: Horizon’s Campanelli Slams CTV’s ‘Extra Step’ to Real Performance
- Read more: Upfronts 2024: AI Hype Crashes Into Measurement Reality
Frequently Asked Questions
What are AI guardrails in ad tech?
They’re predefined rules—business outcomes like revenue or retention—that stop AI from optimizing junk metrics like raw clicks. Think leashes on overeager dogs.
Does Omnicom’s Real ID solve cross-platform measurement?
It unifies identities across channels in clean rooms, promising privacy-safe insights. Promising, but competes with LiveRamp and others—jury’s out on dominance.
Will AI guardrails replace traditional dashboards?
Maybe. Yonosko says dashboards are toast; guardrails push actionable outcomes. But expect hybrids—old habits die hard.
