Jack Dorsey just said the quiet part loud: AI doesn't need permission slips from middle management.
The Summary
- Block's CEO Jack Dorsey outlined a vision where AI replaces middle managers, weeks after cutting nearly half of Block's workforce
- This isn't cost-cutting theater. It's a structural bet that coordination layers are now computational problems, not human ones.
- First major tech CEO to explicitly frame AI as management infrastructure, not just productivity tooling
The Signal
Block just cut half its people. Now Dorsey is explaining why he thinks that's the beginning, not the end. The framing matters here. He's not talking about AI assistants that help managers manage better. He's talking about replacing the management layer entirely with systems that route information, make resource allocation decisions, and coordinate work between teams.
This is the agent economy thesis playing out in real time at a public company with 8,000 employees (well, 4,000 now). Middle management exists largely to solve information flow problems: what's happening, who needs to know, what gets prioritized, how do we allocate budget and people. These are increasingly computational problems. An AI agent can parse signals across Slack, email, project management tools, and financial systems faster than any VP can. It can identify blockers, suggest resource reallocation, and surface decisions that need human judgment without the overhead of status meetings and deck reviews.
What Dorsey is betting is that the actual judgment calls, the ones that need human intuition and values alignment, can flow directly from senior leadership to individual contributors. The stuff in between, the coordination tax, can be automated. Whether that's true at scale is the experiment Block is now running. But the fact that a major fintech company is restructuring around this assumption changes the conversation. This isn't a future-of-work white paper. It's a live deployment.
The Implication
Watch how this plays out over the next two quarters. If Block's velocity increases and product quality holds, every tech company with bloated middle layers will have their own "do we actually need all these people?" conversation. The counter-signal to watch: whether the remaining employees burn out from decision fatigue, or whether the AI systems actually handle the cognitive load Dorsey thinks they will. This could be the blueprint for Web4 organizations, or it could be a very expensive lesson in what humans are still necessary for.
Source: Bloomberg Tech