AI Briefing
KO

AI-Native Organization (Jack Dorsey, Twitter Founder)

·2026.05.21 09:57

Key point

Block CEO Jack Dorsey cuts staff by 40% and flattens hierarchy in an AI-centered organizational overhaul

Details

In a conversation with Sequoia Capital's Roelof Botha, Block (formerly Square) CEO Jack Dorsey revealed an experiment to rebuild the company as 'a single intelligence.' At the core of this change, carried out alongside a recent 40% staff reduction, is replacing traditional management hierarchy with AI.

Redesigning the Organizational Structure

  • Roles simplified into three types: IC (builder/operator), DRI (directly responsible individual for customer outcomes), and player coach
  • The hierarchy from CEO to the front line, currently 5 levels, will be reduced to 2-3 levels this year, ultimately aiming for a structure where every employee reports directly to the CEO
  • An AI layer built on top of all company outputs—Slack, email, code, meeting notes—is designed so that anyone can query it as if conversing with 'the company itself'

Background of the 40% Layoffs

The decisive factor was the judgment, in December 2024, that coding models had matured enough to handle even legacy codebases, with hallucination issues greatly reduced. Starting from the question, "If we were building this company for the first time today with this tool, what would it look like?", they calculated the minimum headcount needed to maintain services, comply with regulations, and fulfill growth commitments.

Turning the Company into a Mini AGI

Dorsey is modeling the entire company like a single artificial general intelligence, implementing a structure that allows real-time queries in board meetings, earnings calls, and customer conversations. Whereas meetings were run with slides just two months ago, now every member brings an actual working prototype.

Dorsey's core argument is that "hierarchy was merely a means of conveying information, and once that means can be replaced by superior technology, any company that fails to rebuild its organization risks becoming a knockoff of the frontier labs." However, he himself acknowledges that this model may work better for businesses with deep customer signal, making it difficult to apply to every company.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.