Ad Tech Briefing: Accountability in adland's agentic era?

"For all the enthusiasm around "agentic" systems, which can plan, optimize, and measure the effectiveness of media campaigns with minimal human intervention, i.e., realize the headcount savings currently wreaking havoc on adland's workforce, one topic that will thrill accountants less so keeps arising: accountability."

"Namely, if AI agents are increasingly making marketing decisions, who ultimately owns the consequences when those systems fail, bias outcomes, optimize toward the wrong goals, or simply produce mediocre work at scale?"

"Speakers repeatedly referenced the industry's tendency to automate flawed processes rather than redesign them, with former IAB U.S. CEO Randall Rothenberg presenting findings that 95% of generative AI pilots fail to deliver measurable profit-and-loss impact, and another forecast that 40% of agentic AI projects will fail by 2027."

""The future of creativity will not be automated," she said, taking aim at some of the shortcomings of programmatic advertising. "Programmatic is really good at attributing value to people who were already on their way to you," she said, arguing the industry had spent years prioritizing measurable efficiency over originality and persuasion."

Agentic advertising systems can plan, optimize, and measure media campaign effectiveness with minimal human involvement, but accountability remains a central concern. When AI agents make marketing decisions, responsibility for consequences such as system failures, biased outcomes, wrong-goal optimization, or low-quality work becomes unclear. Attendees debated how the industry often automates flawed processes rather than redesigning them. Findings were cited that most generative AI pilots fail to show measurable profit-and-loss impact, and that many agentic AI projects are expected to fail within a few years. Optimization culture was criticized for prioritizing measurable efficiency over creativity and persuasion, with AI-generated creative described as optimized for adequacy rather than originality.

Read at Digiday