Exclusive interview: Google’s first chief decision scientist says leaders misunderstand AI

Cassie Kozyrkov, Google’s first-ever chief decision scientist and founder of consulting firm Kozyr, is best known for championing decision intelligence—a field that helps leaders move beyond instinct and data overload toward clarity and structure. In an exclusive interview for Performance Magazine’s AI Edition, she argues that many leaders still get AI wrong by focusing on automation rather than decision quality.

Kozyrkov draws a sharp line between tool use and automation. She points out that the former empowers employees to use AI systems like ChatGPT or Claude to do better work, while the latter involves handing off decisions entirely to machines. Leaders often conflate the two, she warns, launching AI projects without asking whether they truly understand what outcomes they’re pursuing.

Instead of starting with data, Kozyrkov believes that a good AI strategy starts with intent. “The answers are cheap now,” she says. “It’s the questions that are valuable.” For her, prompting is a leadership skill—one that requires clarity about what success looks like and how to recognize it. Off-the-shelf metrics, like accuracy or speed, often miss the point if they’re not tied to the organization’s core priorities.

Kozyrkov also emphasizes the importance of technical fluency at the leadership level—not necessarily coding expertise, but enough knowledge to manage complexity and ask the right questions. She warns that a “non-technical leader who ignores the technology is a hazard” and that trust, especially in external vendors or automated systems, should always be earned—not assumed.

The conversation spans real-world examples, like AI-driven parking lot metrics and automation misfires, to broader themes of governance and accountability. Kozyrkov calls for what she terms a “new breed of leader”: someone who understands that in a world flooded with tools, what matters most is judgment.

Read the full interview in Performance Magazine – Print Edition, Issue No. 33 (The AI Edition), now available as a free digital download via the TKI Marketplace. Print copies can also be ordered through Amazon (standard printing and shipping fees apply).