Insurers Are Building a Layer of AI Accountability in Ad Tech

Joining Louie on stage at Programmatic IO were Nicholas Godlove, director of AI legal affairs and global data privacy at Yum! Brands (owner of Taco Bell and Pizza Hut), and Allan Parker, director of innovation at agency network BBDO North America. They discussed how companies are managing the growing role of insurance providers in AI compliance amid an increasing number of international laws and regulations.

According to Godlove, implementing company-wide policies is "extremely difficult" due to the many types of AI technologies and use cases in the advertising industry: from AI-generated creatives and workflow tools to agentic media buying.

Corporate guardrails

That's precisely why media and technology companies, including brands and agencies, are building their own "guardrails" for the safe and profitable use of AI.

From an agency perspective, "one of the most important things when implementing AI tools is making sure people understand what's available and how to use it," said BBDO's Parker. One tactic is a team of experts who train clients and employees on AI use cases while simultaneously ensuring oversight of the tools' correct operation.

BBDO has an "implementation and support team" with specialists for specific products and scenarios, Parker said. For example, the agency regularly tracks the effectiveness of agentic workflows.

Yum! Brands operates in a similar way: employees must clearly understand which AI products are available to them and what tasks they solve. The company has a dedicated AI team overseeing the legal aspects and safety of implementation. It also ties all of the company's brands together with a unified AI usage policy and standard contracts with AI vendors, including OpenAI.

"The rules we've built over the past two to three years have become increasingly strict," Godlove noted. Today, Yum! Brands takes a more conservative approach: establishing clear guidelines and standards early on, letting employees easily turn individual AI features on and off, and informing clients about the fact that AI is being used.

The requirement for auditability

A cautious approach helps, but what do you do when something goes wrong? AI products and workflows must be fully "auditable," Louie insists — both for training purposes and for determining the cause of a failure.

If an ad campaign misfired in its impressions or missed revenue goals, the agency or tech company should be able to go back and easily find answers: which large language model was activated at the time of the error, what data it generated, and whether the mistake was human or machine.

"No matter which side you're on — agency or company — this level of auditability is really important for building trust in the tools you use," Louie concluded.