Jack Dorsey Proposes Reshaping Manager Roles with AI

Jack Dorsey coauthored an essay with Sequoia partner Roelof Botha titled "From Hierarchy to Intelligence," arguing that AI-enabled systems and smaller teams can replace traditional middle management, per Fortune. Fortune also reports the essay calls for a persistent "world model" of operations and stronger customer signals to substitute for managerial routing of information. In public remarks reported by Yahoo Finance, Dorsey said at a JPMorgan conference that Block is currently about five layers deep beneath him and that he would like to reduce that to two to three layers by the end of the year. Yahoo Finance and Business Insider additionally report Dorsey said on a podcast that his "ideal case" would have all 6,000 Block employees reporting directly to him. Block published a company post summarizing efforts to "organize as intelligence," which aligns with the essay's themes.

Industry reporting frames the proposal as a convergence of two trends: adoption of AI copilots for knowledge work and ongoing organizational flattening. Companies are increasingly experimenting with AI tools that surface context, execute repetitive tasks, and centralize shared knowledge. Observed patterns in similar transitions: firms that push knowledge into searchable, versioned systems aim to reduce handoffs and the need for managers to act as information routers, but they also face challenges in capturing tacit context and ensuring decision traceability.

The proposal follows Block's workforce reduction in February that cut roughly 40% of staff, as reported by Fortune and Yahoo Finance. Public coverage places Dorsey's essay and remarks alongside other tech companies trimming middle management or layers, including recent moves reported at Amazon and Meta. For practitioners, the core significance is operational: replacing manager-mediated alignment with system-mediated alignment depends on high-quality data capture, persistent "world models," and tooling that embeds intent, provenance, and rationale into workflows.

For practitioners: Building the type of company Dorsey and Botha describe requires durable engineering investments rather than incremental UI overlays. Teams would need provenance-aware data stores, richer metadata on decisions and experiments, and tooling for ranking and surfacing authoritative signals to humans and models. Observed patterns in organizations that attempt similar shifts include heavier investment in MLOps, knowledge-graph style indexing, and tooling for human-in-the-loop verification. Those capabilities change where and how data science and ML engineers spend time: more on robust telemetry, model explainability, and integrations that feed an operational "world model." This paragraph is industry analysis, not a claim about Block's internal roadmap.

• •Adoption metrics: whether Block or peers publish concrete measures of speed-to-decision, error rates, or time-to-delivery after flattening layers.
• •Tooling signals: open-source or vendor announcements for "world model" primitives, decision-logging APIs, or provenance frameworks aimed at replacing managerial routing.
• •Organizational experiments: case studies from companies that scale manager-to-engineer ratios dramatically, including attrition, onboarding time, and employee satisfaction, as reported by neutral outlets.

Editorial analysis: Dorsey and Botha's proposal crystallizes a debate about the limits of hierarchy in knowledge work and the plumbing required to run large organizations as intelligence layers. Realizing that vision is primarily an engineering and product challenge for ML/DS teams: it requires durable, auditable, and searchable decision data, not just more LLM copilots. The broader consequences for talent structures and workplace experience will be visible only after organizations publish measurable outcomes from large-scale experiments.