You’ve had this conversation. Your manager asks for a “quick update” on your project. Three hours later, you’ve written a report, prepared slides, and sent a Slack summary — all for a meeting that could have been an email. But the email wouldn’t have justified their existence.
ByteDance CEO Liang Rubo just sent a company-wide letter demanding that leaders produce “substantial output” and get closer to the front lines. On paper, it sounds like a noble attack on managerial bloat. In practice, it will make your work life worse. Managers who can’t be proven wrong are managers who will never be held accountable.
Let’s start with what Liang actually said. He argued that in the AI era, the old model of middle managers as information relays, approval gatekeepers, and process enforcers is dying. AI can already summarize meetings, generate reports, and approve standard requests. So what’s left for managers? Real, tangible output that moves the business forward. High goals. Being in the trenches.
Sounds great. But here’s the problem: “substantial output” is a phrase so vague that every manager will interpret it as “make my reports work harder.” The most dangerous phrase in corporate speak isn’t ‘synergy’ – it’s ‘substantial output’.
I’ve seen this movie before. A former colleague of mine at a state-owned enterprise watched their new chairman — introduced as having “financial background” — spend his days in meetings about meetings. His “financial background” turned out to be two years as a cashier in the 1990s, followed by two decades in party affairs. He wasn’t evaluating anything; he was delegating evaluation to his subordinates while claiming credit for their analysis. That’s the pattern: when output is undefined, managers define it as everything their team does.
The deeper issue is what I call the falsifiability test. A real manager’s work should produce something that can be proven right or wrong. Did the algorithm prototype run? Yes or no. Did the audit find a material misstatement? Show me the adjustment. But most managers operate in domains where their decisions cannot be falsified. Did you need that approval step? Without it, maybe something bad would have happened — except we’ll never know. That’s the perfect protection racket.
ByteDance is huge, with dozens of business lines from TikTok to enterprise software. Liang can’t personally audit each leader’s output, so he uses a fuzzy word. And that word will be hijacked. When you can’t measure managers, they measure you instead.
Think about what will actually happen. A manager who wants to look productive will ask their team for more granular reports, more frequent check-ins, more detailed documentation — all under the banner of “output.” They’ll use AI to automate the collection of this data, then present it to leadership as evidence of their own effectiveness. The employee who was already drowning in meetings now has to feed the machine that reports on how hard everyone is working. AI becomes a whip, not a ladder.
This isn’t theory. I’ve lived it. A previous boss required nightly write-ups of every client conversation, including the exact phrasing I used. Then he’d ask me to highlight the key points — for him. Then he’d copy those highlights into an email to his boss with no attribution. His “output” was my output, repackaged. And he got promoted.
So where does that leave us? Liang Rubo is right that process-heavy managers are redundant. But he’s wrong to think that asking for “output” will fix it. Without falsifiable metrics — something a manager can be objectively wrong about — the system will reward managers who are best at extracting work from others and claiming it as their own. Clarity without accountability is just a new layer of paperwork.
What would actually work? Demand that every manager, at least in one domain, produce work that can be tested. A prototype that runs. A forecast that can be compared to reality. A strategic bet that either pays off or doesn’t. No more “we align” or “we synergize.” Show me a decision you made, and show me how we’ll know if you were right. Until then, every corporate reform is just a fresh coat of paint on a leaking ship.
FAQ
Q: What's the practical implication for employees after this ByteDance policy?
A: Expect more meaningless work, not less. Managers will demand extra reports and check-ins to demonstrate their own 'output', using AI to automate surveillance. Your workload increases while theirs becomes a dashboard of your labor.
Q: Is this just a ByteDance problem, or does it apply to all large companies?
A: This is universal in any organization where managers' value can't be directly measured. The lack of falsifiable metrics is endemic in corporate hierarchies. ByteDance is just a high-profile example of a systemic failure.
Q: What's the real solution to managerial bloat in the AI era?
A: Require every manager to produce falsifiable work: a prototype, a forecast that can be checked against outcomes, or a strategic decision with clear success criteria. Without that, any reform will be co-opted to justify existing hierarchies.