The Interactive Advertising Bureau's Gabilan has put a blunt message in front of the ad-tech trade: agentic AI won't paper over ad measurement's existing problems. It will expose what he calls the "old cracks" already running through the system.

That framing matters because it cuts against the current mood. Most of the industry chatter around agentic AI treats it as an upgrade — smarter buying, faster optimization, fewer humans in the loop. Gabilan's warning flips that: if software agents start acting on campaign data at machine speed, every weakness in how that data gets measured shows up immediately, and at scale.

The venue was Yahoo Finance, which carried the headline. The substance is the part practitioners should sit with.

What does this actually change for operators?

Here's the practical question. If you run buying, planning, or measurement for a brand or agency, what moves for you?

Plenty, if the prediction lands. Agentic AI systems don't just read reports — they act on them, adjusting spend and creative placement without a human signing off on each step. That works only as well as the measurement feeding it. When humans review dashboards, they apply judgment and can quietly discount numbers they distrust. An agent applies the numbers as given. Garbage in, budget out.

So the "old cracks" Gabilan points to stop being background noise. Gaps in cross-channel attribution, inconsistencies between vendor panels, the usual disputes over viewability and reach — these become live inputs an agent optimizes toward, flaws included.

Why the IAB's voice carries weight here

The IAB sets standards and frameworks that much of the digital ad ecosystem leans on, from measurement guidance to compliance programs. When a named executive from that body says measurement has cracks agentic AI will expose, it reads less like vendor FUD and more like a standards-body flag.

It also signals where the IAB's attention is heading. Expect the trade group's measurement work to tilt toward questions raised by automated, agent-driven decisioning — not just whether ads were measured, but whether the measurements are robust enough for machines to act on autonomously.

The skeptic's read

Strip out the buzzword and the argument is old-fashioned: automation amplifies whatever it's built on. Feed an agent clean, consistent measurement and it compounds the gains. Feed it the patchwork the industry runs on today — where the same campaign can score differently depending on who's counting — and it compounds the errors, faster than any human team can catch.

That's not an argument against agentic AI. It's an argument for fixing measurement before, or at least while, the agents arrive.

For operators, the takeaway is straightforward: audit what your measurement stack actually delivers before you wire it into anything autonomous. The cracks are already there. Agentic AI just makes them impossible to ignore.