Block replaces management with AI agents, cuts 40% of staff
- Jack Dorsey is restructuring Block as a ‘mini-AGI,’ eliminating middle management to prioritize AI-driven information flow.
- The company slashed development headcount by 40% after agent productivity made traditional software teams obsolete.
- Internal AI agents now autonomously write code and generate unique app interfaces for each user.
Jack Dorsey is dismantling the corporate hierarchy at Block, arguing the 2,000-year-old management model is a bottleneck for information. His new structure has three roles: Individual Contributor, Directly Responsible Individual, and Player-Coach, eliminating traditional middle management. The goal is to turn the company into a collective intelligence, where every Slack message, email, and code commit feeds an AI model that understands the organization better than any human executive.
The shift is already operational. According to Block executive Owen Jennings on The a16z Show, the historical link between employee headcount and productivity “basically broke” in late 2024. The company responded by cutting 40% of its staff, specifically in development, replacing 14-person feature teams with squads of one to six people augmented by AI agents.
Coding is no longer a manual task. Block uses an internal agent harness called Goose and a tool named Builder Bot, integrated into Slack, to autonomously write, test, and merge code. Jennings describes managing 10 or 20 agents in the background, with humans acting as editors for the final 10% of a feature. Designers and product managers now ship their own code directly to production, compressing development cycles from months to weeks.
The endgame is dynamic, personalized software. Block is deploying AI like MoneyBot for Cash App to generate custom interfaces for each user on the fly, moving beyond static apps. This creates a massive QA challenge but represents a fundamental break from the one-size-fits-all software model.
Owen Jennings, The a16z Show:
- There's been this correlation between the number of folks at a company and the output from the company for decades and decades.
- I think that basically broke.
- We're not writing code by hand anymore. That's over.
Dorsey’s vision, outlined on Long Strange Trip: CEO to CEO, treats the company itself as a training dataset. The AI-modeled ‘mini-AGI’ is meant to provide any employee direct access to the organization’s collective intelligence, making board meetings strategic rather than operational. The experiment is a direct challenge to the law of software headcount, betting that deep data insights, not human scale, will be the new corporate moat.