Analysis Group Affiliate Lee Tiedrich Charts a Path Toward Closing the AI Evaluation Gap

August 12, 2026

Current AI evaluation methods often fail to reflect real-world performance. The disparity between the power of new AI systems and our ability to measure their effectiveness and safety is known as the AI evaluation gap. Closing the gap could boost trust, security, adoption, and policy clarity while lowering costs and barriers to competition. In an OECD.AI blog post, affiliate Lee Tiedrich proposes a five-step roadmap toward bridging this divide, including balancing standardization with customization, testing AI applications throughout their life cycles, and tailoring evaluations of those tools to different actors across the AI value chain.

“Lee gets at something the field has been circling for a while: Evaluating AI tools is an ongoing discipline that has to evolve alongside the technology itself.” – Jimmy Royer

The full article is available on OECD.AI's blog, The AI Wonk: A five-step roadmap to closing the AI evaluation gap

Ms. Tiedrich has also been active elsewhere in the AI policy conversation, appearing on NPR's All Things Considered to discuss AI regulation and presenting at Ai4 2026, North America’s largest AI conference.