From managing people to managing AI: The leadership skills everyone needs now | Julie Zhuo (Facebook VP, Sundial CEO, The Making of a Manager author)

The conversation opens with a familiar prediction: as AI becomes part of everyday work, organizations may flatten and more people may return to individual-contributor work. Lenny Rachitsky points to the apparent reduction of middle-management layers, while Julie Zhuo argues that the trend should not be confused with the disappearance of management. Her definition is outcome-oriented. A manager starts with a north star and a vision, then assembles the resources needed to achieve them. Those resources used to be mostly people; now they also include models and the tools those models can use. The management problem therefore shifts from supervising a fixed hierarchy to choosing and coordinating a changing set of capabilities.

Zhuo's metaphor for this environment is a willow tree. Managers still have to absorb disruption, but the pace of change makes flexibility as important as endurance. The point is not to preserve an old organizational shape. It is to remain steady about the outcome while adapting the people, models, tools, and processes used to reach it. This leaves an important qualification to the flattening thesis: fewer formal managers may mean more responsibility for management work at every level, rather than less management work overall.

Zhuo connects her earlier work leading design at Facebook with her current focus on data and analytics through a compact division of labor: data helps diagnose a problem, while design treats it. In this account, metrics do not dictate what a company should build. They help reveal where growth, behavior, or a product experience is breaking down; judgment and design are still needed to decide what response makes sense. She is skeptical that many fast-growing companies use data well, because rapid growth can temporarily be sustained by instinct and enthusiasm. The limitation is temporal: those habits can look sufficient until growth stalls, at which point a company needs better diagnosis.

The episode is also a return to the ideas in The Making of a Manager. Rachitsky introduces Zhuo as the former head of design for the Facebook app, the author of that book, and the founder of Sundial, a company focused on using AI for data analysis. He presents the range from a large company to a small startup as central to the discussion: Zhuo has seen both the coordination demands of a very large organization and the decision-making needs of a young company. The setup makes the interview less about a generic forecast and more about which management habits survive a change in scale, technology, and role.

Zhuo says the book exceeded her expectations, but its original purpose was personal as much as commercial. Writing down what she had learned was a way to become a better manager and to make feelings that had seemed private, including impostor feelings, easier for readers to recognize as normal. That origin matters for the AI discussion. The book is presented as a record of thinking and practice rather than a claim that management can be reduced to a timeless checklist. Its continued usefulness comes from naming the uncertainty managers experience while they are learning the work.

The early exchange about the podcast's growth and the guests' long-running relationship supplies context rather than evidence for the management thesis. It shows a conversation revisiting older management ideas after three years of technological and professional change. The warmth of the reunion, including their discussion of the podcast's evolution from audio toward video, also establishes that the episode is exploratory: the participants are testing how familiar principles map onto a new setting rather than presenting a finished operating manual.

Zhuo maps traditional management onto agentic systems through three linked levers. First comes talent or capability: a manager needs the right mix of strengths, whether those strengths belong to people or models. Second comes purpose: everyone and everything involved needs to understand the goal and what its contribution is meant to achieve. Third comes process: the parts must fit together in a way that produces the intended result. The analogy is practical rather than anthropomorphic. Models may have different strengths and recognizable tendencies, but the manager's job remains to match a capability to a purpose and then design the workflow around that match.

The translation is not complete or automatic. Better models may take on more of the work of figuring out how to proceed, but Zhuo still emphasizes context, high-level instruction, and process decisions. The human remains responsible for choosing the goal, supplying the information that makes the goal intelligible, and deciding how tools interact. In that sense, using agents well resembles managing a team with unfamiliar specialties: competence comes from learning what each participant can do, not from assuming that one general instruction will work everywhere.

Rachitsky tests the analogy against a conventional description of a manager's job: build a team, support members' career goals, and create efficient processes. He notes that agents do not need career development in the human sense. Zhuo accepts that difference, but keeps the conversation deliberately light by observing that people may still choose to treat agents kindly in case future systems become more capable. The exchange separates the human-care part of management from the coordination and goal-setting parts that Zhuo believes transfer directly.