It’s the idea that at large companies the incentive to deliver something real to the customer becomes a small fraction of work, and is replaced by projects that can sound very meaningful on paper/jira but aren’t. Here’s a pretty good take on it.
One of the biggest lies of work theater is “If we aren’t succeeding, we need to stop asking questions and work harder!”
This advice, which boils down to “Think less, act more” is good advice when you have product-market fit and know exactly what to build. I’d hazard that only a tiny percentage of the time does a company have one key project that’s make-or-break for them. When you don’t know exactly what to build you need to do the opposite: slow down, ask more questions, and solicit input from a wider range of people.
The AI era promises a much faster 0->1 transition (when you think you know what to build). But it does nothing to help with where to strategically go from there (how do we grow this thing?). In fact, it makes it exponentially easier to play out the theatrics of progress (look at all this motion! So many tickets! So many changes!) while treading water quarter after quarter.
When I look at big companies that failed (yahoo et al), I can’t think of a single time where the cause of failure was that they were building the right thing but were 20% too slow. Almost every time the failure is focusing on the wrong thing. Often there are plenty of people (employees, customers, randos on forums) calling out the directional failings of the company.
Thus, I don’t see AI fundamentally helping large companies, because speed was never the issue.
In Recursive Middle Manager Hell Raemon proposes the primary factor of company waste is middle-management-layers. I’ve worked at a dozen tech companies, and this squares pretty well with what I’ve seen – it’s not the only factor that matters but it’s one of the biggest.
The anti-pattern of building-the-wrong-thing doesn’t get solved by a 2x or 3x speed-up. Thus my prediction (and experience) is that companies with more than 4 levels of managers (including executives) are going to see small or negative improvement from agentic work. The reason I could foresee negative is that AI expands the capacity of the most optics-driven career-oriented engineer to create drag on everybody around them (tech debt).
Conversely I think that companies with three levels of management or less will see significant acceleration, as fewer people can get more done. They will be able to achieve 0->1 more quickly, and build out numerous integrations and basic features.
This is probably a good thing. Competition forces a superior user experience, and large lazy incumbent companies (be it Yahoo in 2010 or Meta in 2026) should raise the bar on what they deliver or be replaced.