Delegating to AI is not delegating to people: the weight of oversight remains
Handing projects to AI seems to free you, but oversight and responsibility stay on your shoulders—making the old ‘let it go’ wisdom suddenly obsolete.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Coding brought flow; reviewing AI work doesn't replace it
Engineering transformed in two years: writing code gave way to talking to AI, waiting for agents, then reviewing their work. Molly validates this as genuine grief, not whining. Many engineers tell her what they truly miss is the flow state of sitting down to focus on pure code work, not their new role of reviewing and correcting AI output. Paddling felt generative; steering alone feels hollow. The people who loved paddling don't automatically become people who love steering, but we're told it's a promotion, not a loss.
— Molly GrahamSmaller teams mean less human connection, more AI interface
Shrink an engineering team and each person interfaces with more AI agents but fewer human colleagues. Fiona Fung observed that her team now spends entire days talking only to Claude. Molly notes this is a trade-off organizational designers must face: you can extract efficiency gains by cutting middle management, but the cost is isolation and low morale—and those don't produce best-quality work. Research also shows people at small teams and small companies are visibly happier than those at large ones. The efficiency equation doesn't account for that.
— Molly GrahamMost ‘AI layoffs’ are hiring chaos rebranded, not displacement
Molly is skeptical of the ‘AI layoffs’ narrative now ubiquitous. Most aren't actually AI replacing roles—they're companies with poor management that overhired and now relabel cuts as AI-driven to boost share price and dodge admitting hiring failures. There's not yet substantial data proving AI is replacing jobs at scale, but the fear narrative alone already blocks employees from embracing change with hope. They're just in self-protection mode.
— Molly GrahamBurnout rises 11 points in one year as AI multiplies demands
Lenny's consecutive annual industry surveys show ‘feeling burned out’ jumped from 44% last year to 55% this year—a full 10-point rise. Employees are asked to accomplish more via AI, earn roughly the same wage, and watch sibling teams ‘move faster’ with anxiety. But the same data shows half the workforce is in the happiest stretch of their careers. The happiest group: those who genuinely feel ‘multiplied’ by AI, working in smaller teams with real autonomy.
— LennyAI delegation doesn't erase accountability—it concentrates it
Molly's advice for 13 years was ‘delegate completely and stop worrying.’ But delegating to AI differs fundamentally from delegating to people. Regardless of capability, you hold full accountability for AI output. Oversight costs don't evaporate. Outwardly you handle more; inwardly you're carrying all that AI accountability. You have to oversee not just AI but often a bigger roster of people too. That likely explains part of the burnout spike: your mental load climbed instead of falling.
— Molly GrahamYour management ceiling is 10 to 12, whether leading people or AI
Molly has long held that one manager can effectively lead roughly 10 to 12 people—she's pushed past that and the results were poor. She thinks this ceiling applies equally to managing AI agents. Count agents in your direct reports, and many people have already blown past the limit. Delegating to AI uses the same skills as managing people: give context, correct errors, coach. But AI resembles an unpracticed summer intern who needs more rounds of guidance. That makes ‘everyone's a manager’ an unavoidable reality.
— Molly GrahamSome work demands judgment—don't hand it to machines
This is Molly's largest revision to 13 years of doctrine: don't outsource everything. Some ‘Lego blocks’ should stay in-house—work requiring judgment, defining ‘what good looks like,’ and relationship-based tasks. She cites a CEO copying an AI-written strategy memo and sending it company-wide as a case study in how the organization signals that ‘thinking and judgment are outsourceable.’ That signal is corrosive. The parts humans genuinely excel at shouldn't defer to machines still operating at summer-intern caliber.
— Molly GrahamThe practice that compounds: ask if AI can help before you start
Lenny says the highest-leverage skill to cultivate now isn't mastery of a specific tool—it's asking ‘can AI help me with this?’ before you begin anything. Sounds basic, almost awkward, but it mirrors meditation's ‘space between stimulus and response.’ Once you cement the habit, you keep spotting new applications for AI and widen the gap between your practice and everyone else's. That's the compounding edge.
— LennyIn their own words · checked verbatim
the job used to be rowing. And now it's like, steering, steering. And I was like, I feel like there's a lot of people out in the world right now that are like, I don't want to fucking steer.
Lenny20:40
I want you to pour every single thing that you know into this employee. And then they're going to take your job in six months. Like who the fuck wants to do that?
Molly Graham25:53
I have a lot of beef to pick with all the AI branded layoffs out there because they're bullshit. Like they're not about AI. Like they're about badly run companies slapping an AI label and getting some share points from that versus saying, whoops, we hired too many people.
Molly Graham26:53
I feel like we need a change from like this super intelligent being that's better than all of us at our jobs to like, no, it's just a fucking intern, man.
Molly Graham37:11
delegating something to a robot is not the same as giving it to a human. Because you cannot get rid of the oversight, right?
Molly Graham44:42
You individually are phenomenal at a set of things like do not outsource that. Like you can enable it. You can amplify it to use your data, but don't outsource it to these weird robots that, you know, are effectively summer interns.
Molly Graham1:05:23
sometimes you need to throw a funeral for things. Do you know what I mean? And I was like, yeah, man, sometimes you need to, the funeral is there for a reason.
Molly Graham1:07:23
I feel like the biggest skill to build right now is when you're about to do something to ask yourself, can AI help me with this thing?
Lenny1:13:40
Figures
| Burnout increase year-over-year | 44% to 55% | 29:55 |
| Engineering rework code volume increase | 8x | 39:18 |
| Manager team-size ceiling | 10-12 | 47:48 |
Glossary
- Centaur
- Humans direct AI—the model where people decide, AI executes. Its inverse, ‘reverse centaur,’ is AI directing humans, like food-delivery drivers scheduled by algorithms.
- AI slop
- Copying AI-generated content without filtering or refinement—low-quality output that leaves others to clean up.
- Human sandwich
- Organizational structure: humans set direction and vision at the top, AI generates heavily in the middle, humans handle QA at the bottom.
How to listen
Engineers and product managers actively delegating work to AI, plus managers worried about team morale and role transitions.
Around 55:07, there's a section with obvious audio/transcript glitches and repetition; it's low on substance and worth skipping.