Agentic AI in advertising is real, but autonomous campaigns are still a work in progress

The advertising industry is in the middle of a shift in how campaigns are planned, bought, optimized and measured. This shift is driven by AI systems that don’t just recommend actions but execute them.

Buzzwords like agentic AI continue to be floated with the idea that campaigns are being fully built and delivered by autonomous AI systems, but that doesn’t quite align with reality today. 

“The hype is that agentic AI is already running media end-to-end autonomously. It’s not,” said Jamie Finstein, vice president of the IAB Media Center.  “What’s real is that it has the potential to improve workflow. But the infrastructure required for true autonomy — standardized data, interoperable systems and governance frameworks — is still being built.”

Finstein said the industry is in a transitional phase, one where the technology is advancing faster than the structures needed to support it.

That assessment is echoed in a new whitepaper from the Interactive Advertising Bureau, “AI-Powered Video Outcomes: Agentic AI,” which examines how autonomous systems are reshaping video advertising and where the gaps remain.

Control moves upstream

One of the changes Finstein described is a restructuring of how humans interact with campaign management. Rather than managing campaigns step by step, advertisers are increasingly defining goals, constraints and guardrails before a system begins operating.

“Instead of managing campaigns step-by-step, humans are defining goals, constraints and guardrails upfront — things like outcomes, budget limits, audience parameters and privacy requirements,” Finstein said.

That change carries new forms of risk.

Advertisement

In an environment where AI agents act on predefined inputs, poorly defined parameters can produce compounding errors at speed.

“If the inputs aren’t clearly defined, the outputs won’t be reliable, and in an agentic environment, those errors can scale quickly,” Finstein said.

Measurement shifts from reporting to decision-making

The IAB whitepaper reported that nearly 70% of analytics teams are already scaling AI, underscoring how quickly measurement is moving from a retrospective function to an active one. Data is no longer used solely to evaluate past performance — it feeds directly into real-time optimization.

That shift raises questions about trust and transparency.

“Advertisers need visibility into how data is normalized, what methodologies are applied and why decisions are made,” Finstein said. “Otherwise, it becomes a black box.”

Finstein also pointed to the risk of AI-generated outputs that are inaccurate or inconsistent, sometimes called hallucinations, which she said can erode confidence in the systems if left unaddressed.

The IAB has been working on standardized frameworks through its Measurement, Addressability & Data Center and a program called Project Eidos, both aimed at establishing consistent, validated signals that support AI-driven systems.

Small errors don’t stay small

The whitepaper emphasized that without defined guardrails, the same speed that makes autonomous systems attractive also makes errors difficult to contain.

“In agentic systems, small errors don’t stay small — they compound,” Finstein said. “So guardrails need to be both technical and organizational.”

On the technical side, that includes budget constraints, frequency caps, brand safety filters and optimization thresholds. On the organizational side, Finstein pointed to decision logging, audit trails and escalation protocols that allow teams to trace what happened and intervene when needed.

Finstein said the industry should balance speed and oversight by allowing autonomous systems to optimize in real time while humans retain control over objectives and constraints. Accountability, she added, depends on assigning clear ownership, logging decisions for audits and escalating selected actions to human review.

Creative at scale, with caveats

Agentic AI also opens the door to personalized, multi-version video creative produced at scale.

Advertisement

The whitepaper noted that this capability introduces a tension between personalization and creative differentiation. More versions of an ad do not necessarily mean more distinct creative ideas.

Finstein framed the near-term value in practical terms.

“The most immediate opportunity is unlocking scale,” she said. “While it may not be the most exciting conversation, applying gen AI to automatically optimize creative to meet format and platform specifications can materially expand usable inventory while delivering a smoother, more relevant ad experience for consumers.”

The whitepaper flagged privacy and governance as areas where progress has not matched the pace of technical development. As AI agents access and act on continuous data signals, questions around data usage, permissions and accountability remain open.

The IAB outlined a set of recommendations in the report, including standardized measurement practices, defined guardrails for autonomous execution and greater transparency into how AI-driven decisions are made. The organization positioned these steps as necessary for moving agentic AI from an experimental phase to broader operational use.