Home

 › 

Economy

 › 

AI Just Turned Every Worker Into a Manager. Molly Graham Says Many of Them Actively Chose Not to Be One.

Ink illustration: a long gray reception counter lined with a row of small silver service bells, one orange bell just rung, with an empty office chair behind the counter.

AI Just Turned Every Worker Into a Manager. Molly Graham Says Many of Them Actively Chose Not to Be One.

Quick Read

  • Most workers who avoided management did so deliberately, and AI simply overruled that decision without asking.
  • Adam Grant added AI agents to boost productivity. What he got felt nothing like what the pitch promised.
  • Graham makes two arguments about AI that directly contradict each other, and he never resolves the tension between them.
  • Just released. Our analysts combed the entire stock market and named the ten best stocks to buy right now. The report is free. Enter your email and see if any of your stocks made the cut.

Where we heard itLenny’s Podcast, “The grief, loneliness, and burnout sweeping through the tech industry right now.” Released Sept. 27, 2026, and reposted Oct. 4 in TED’s WorkLife feed. Speaker: Molly Graham, host of WorkLife, in conversation with Lenny Rachitsky. Listen at Lenny’s Newsletter

Most coverage of AI at work asks whether the tools will replace people. Molly Graham, a veteran technology operator with nearly 20 years in the industry, offers a sharper observation. In her telling, AI handed every worker a set of direct reports. By her account, assigning to an AI agent is a form of management.

Graham made the case on Lenny’s Podcast, hosted by Lenny Rachitsky, in an episode released September 27, 2026. TED reposted it on October 4 in the feed of WorkLife, the TED show Graham now hosts after taking it over from organizational psychologist Adam Grant. Quotes below come from the episode audio.

Quote: “It is the difference between managing junior employees and managing senior employees.” Molly Graham, host of TED’s WorkLife, on Lenny’s Podcast
My Investing News

Assigning to AI Demands the Same Skills as Managing People

Graham holds that handing work to an AI calls for the same skills as managing humans. “They have less feelings, but you know, it’s all the same skills of like oversight and, you know, giving them context and correcting their work and coaching,” she said.

Every item on that list is a supervisory act: oversight means checking output, context means explaining what good work looks like, correction means catching mistakes, and coaching means improving the next attempt. Someone who prompts an agent, reviews the draft, and sends it back with notes is running a management loop.

Workers Got a Promotion They Never Asked For

Graham’s most pointed argument is about consent. “So many people actively like chose not to become a manager,” she said.

Management is its own career track. Many excellent individual contributors turn it down on purpose: engineers who want to write code, writers who want to write, analysts who want to build models. In Graham’s framing, AI gave those people the job anyway, with no title, no training, and no say in it.

Bar chart: Half of U.S. workers now use AI on the job. Share of U.S. employees who use AI in their role at least a few times a year, by Gallup survey: 21% in Q2 2023, 27% in Q2 2024, 40% in Q2 2025, 45% in Q3 2025, 46% in Q4 2025, 50% in Q1 2026 and 52% in Q2 2026. Weekly use rose from 11% to 30%.
My Investing News

A Hard Ceiling on What One Person Can Hold

Graham says research suggests 10 to 12 direct reports is the most a person can manage well. She has gone over that range herself, with bad results.

Then came her warning: “And if you count robots, like I’m worried for people’s like psychology,” she said, pointing to “the amount you have to hold in your head.”

Her point is that AI agents draw on the same limited attention a human team already uses up. A manager with a full team who adds agents is adding reports, and all of them pull from one pool.

How U.S. managers’ teams look today

Typical team5 to 6Median direct reports, steady even as the average rose
Very large teams13%Share of managers with 25 or more direct reports
Player-coaches97%Managers who also do some individual work
Own work40%Median share of a manager’s time spent on it

Lenny Rachitsky Says His Agents Behave Like Junior Staff

Rachitsky described his own experience. His AI agents ping him all day asking for approvals and corrections, and it made him appreciate a human engineer who can take a rough idea and run with it. Graham compared it to managing junior employees rather than senior ones. “Like AI right now, pretty junior employees,” she said.

That comparison carries practical insight. Junior employees need the most management of any hire: detailed instructions, review, correction, and guidance. Senior hires handle uncertainty and deliver finished work. If agents act like junior staff, adding more agents adds more supervision, the opposite of the usual productivity pitch where automation means less human effort.

Bar chart: The average manager’s team already sits at Graham’s limit. Average direct reports per U.S. manager, per Gallup: 8.2 in 2013, 9.1 in 2019, 10.5 in 2022, 10.6 in 2023, 10.9 in 2024 and 12.1 in 2025, against the top of Graham’s 10 to 12 range.
My Investing News

Graham Argues Fear Is Holding Workers Back

Graham says the promise of AI as a liberating tool depends on the story people tell about it. Fearful coverage claiming AI will eliminate all jobs keeps workers from adopting automation. If people believed new opportunities awaited on the other side, they would delegate freely. Instead, anxiety makes them reluctant to hand anything over.

Her two arguments pull against each other. One says AI creates a real psychological burden. The other says fear about AI keeps workers from using it well. The conversation leaves both claims standing.

Supervisory Work Without Supervisory Training

Neither speaker fully answers the practical question their conversation raises. Using AI well requires management skill. Most workers were never trained to manage, and many chose not to. Organizations handed supervisory work to their whole workforce without the training that normally comes with it. Graham and Rachitsky identified that gap and left it open.

The bottom line

Graham’s view is that working with AI agents is management work, and it has landed on people who never wanted to manage. The risk is overload: she puts the healthy limit at 10 to 12 reports, and the average U.S. manager already had 12.1 people reporting to them in 2025 before counting any agents.

To top