IAB's Gabilan: Agentic AI Exposes Measurement Cracks

Agentic AI — autonomous software agents that plan, buy, and optimize media with minimal human intervention — is poised to stress-test the advertising industry's measurement infrastructure in ways that legacy workflows never did. According to the IAB's Jack Gabilan, speaking in a recent Beet.TV interview, the arrival of AI agents at the center of media buying will not create new measurement problems so much as expose old ones that the industry has tolerated for years.

The core argument: when humans sit between data and decisions, inconsistent metrics, fragmented identity, and incompatible measurement standards can be smoothed over through experience and manual judgment. When autonomous agents are making those decisions at machine speed and scale, those same inconsistencies become systemic liabilities. The "cracks" that human operators routinely paper over become faults that agents will stumble into, amplify, or exploit.

Why Agents Change the Measurement Equation

Programmatic buying has always depended on signals — impression counts, viewability, completion rates, attribution, and audience identifiers — that are generated by a patchwork of vendors using differing methodologies. A human trader can intuit that one platform's viewability figures aren't directly comparable to another's, or that a measurement vendor's attribution window differs from a competitor's. Agentic systems, by contrast, consume these signals literally and optimize against them relentlessly.

That creates a precision problem. If an agent is instructed to optimize toward a KPI, the quality and consistency of the underlying measurement determines whether the agent converges on genuine performance or on measurement artifacts. Garbage-in, garbage-out becomes more dangerous when the "out" is autonomous spend decisions executed across billions of impressions.

For video and CTV specifically, this matters acutely. Measurement in connected TV remains fragmented across walled-garden reporting, SSAI-inserted impressions, and third-party verification layers. Completion rates, ad-pod positioning, co-viewing adjustments, and cross-device identity resolution all vary by platform and vendor. An agent optimizing CTV campaigns across this landscape inherits every inconsistency in the measurement stack.

Identity and Signal Fragmentation

Gabilan's framing lands squarely on a problem the industry has been wrestling with since the deprecation timeline for third-party cookies began: identity fragmentation. Post-cookie addressability depends on a mix of first-party data, authenticated identifiers, universal IDs, and probabilistic modeling — none of which align perfectly across the supply chain.

Human buyers manage this ambiguity through judgment. Agents require deterministic, machine-readable consistency to function reliably. The gap between what measurement systems actually deliver and what autonomous agents need to operate safely is, in Gabilan's view, where the industry's long-deferred standardization debt comes due.

Implications for Standards Bodies and Vendors

The practical takeaway for the IAB and IAB Tech Lab is that agentic AI raises the stakes for interoperable measurement standards. Specifications that define consistent impression counting, viewability methodology, supply-chain transparency (via sellers.json, ads.txt/app-ads.txt, and the supply chain object), and standardized identity signals become foundational rather than aspirational when agents depend on them to make autonomous decisions.

For measurement and verification vendors — DoubleVerify, IAS, Nielsen, Comscore, and others — the shift suggests that machine-consumable, consistently defined metrics will matter more than bespoke dashboards built for human analysts. APIs that expose measurement data in standardized, auditable forms are likely to become table stakes as agentic buying scales.

For publishers and ad-ops teams, the warning is that any weakness in their measurement and signal pipeline — inconsistent VAST reporting, unreliable SSAI impression tracking, or poorly instrumented viewability — will be surfaced faster and penalized more harshly by agents than by human buyers who might have given the benefit of the doubt.

A Forcing Function, Not a New Problem

Gabilan's central point is ultimately optimistic in its framing: agentic AI is a forcing function. The measurement inconsistencies it exposes are not new, but the pressure to resolve them is. The industry has long known that fragmented identity, incompatible metrics, and opaque supply paths undermine programmatic efficiency. Autonomous agents simply remove the human buffer that has made those problems tolerable.

Whether the response is accelerated adoption of IAB Tech Lab standards, broader deployment of clean measurement APIs, or new frameworks purpose-built for agent-to-agent transactions, the direction is clear. As AI agents move from experimentation into production media buying, the quality, consistency, and transparency of the underlying measurement layer will determine whether the industry builds a reliable foundation — or lets old cracks widen into structural failures.

Stay on top of video ad serving and programmatic. Follow Adelerate.