Recently, Jack Dorsey of Block (formerly Square) and Roelof Botha of Sequoia Capital co-published “From Hierarchy to Intelligence,” a sweeping essay that traces 2,000 years of hierarchical organization and argues that AI fundamentally changes how companies should be structured.

• From Hierarchy to Intelligence (Block, Inc.)
• From Hierarchy to Intelligence (Sequoia Capital)

The essay states: “Most companies are focused on AI as a productivity enhancer. Few are focused on the potential of AI to change how we work together.”

As a startup CTO, I found several ideas in this essay that are immediately applicable, regardless of whether you operate at Block’s scale. Here is what matters and what you can act on today.

The Core Thesis: Hierarchy Was an Information Router

The essay begins by redefining what hierarchy actually is.

From the Roman army’s 8-person squad (contubernium) to the Prussian General Staff system invented after Napoleon’s victory at Jena, to the first corporate org chart drawn by Daniel McCallum for the New York & Erie Railroad in the mid-1850s, the history of organizational design is traced through six eras. The conclusion is clear:

Two thousand years of organizational innovation has been an attempt to work around this tradeoff without breaking it.

The “tradeoff” is this: a single leader can effectively manage 3 to 8 people. Narrowing span of control means adding layers, and more layers mean slower information flow. Every organizational experiment of the past century (Spotify squads, Zappos Holacracy, Valve’s flat structure) attempted to solve this but reverted to hierarchy at scale, because no alternative mechanism existed for routing information.

The essay’s argument is that AI can be that alternative mechanism.

Block’s Four-Layer Model

Block is replacing its traditional org chart with a four-layer structure.

1. Capabilities

Atomic financial building blocks: payments, lending, card issuance, banking, BNPL, payroll. No UIs. Focused purely on reliability, compliance, and performance targets.

2. World Models

This is the essay’s centerpiece. Two models work in tandem:

• Company World Model: A real-time understanding of what is built, what is blocked, and how resources are allocated. It replaces the information-routing function that managers traditionally performed
• Customer World Model: A per-customer, per-merchant understanding built from transaction data. Block holds a unique advantage here: Square sees the merchant side while Cash App sees the consumer side of millions of daily transactions

3. Intelligence Layer

This layer composes Capabilities into proactive solutions at specific moments. For example, detecting seasonal cash flow tightening for a restaurant and automatically proposing a short-term loan with adjusted repayment terms.

The most provocative design choice: when the Intelligence Layer cannot compose a needed solution, that failure signal automatically generates the product roadmap. Instead of product managers hypothesizing what to build, the system’s limitations tell you what is missing.

4. Interfaces

User-facing surfaces: Square, Cash App, Afterpay, TIDAL, bitkey. Important, but explicitly positioned as not where value is created.

Three Roles: An Organization Without Managers

On top of this four-layer model, Block defines three roles.

Role	Description
IC (Individual Contributor)	Deep specialists building and operating Capabilities, World Models, the Intelligence Layer, and Interfaces. The World Model provides context, so ICs can make decisions without waiting for approval
DRI (Directly Responsible Individual)	Owns a specific cross-cutting problem or customer outcome for a defined period (e.g., “reduce merchant churn in this segment over 90 days”). Has full authority to pull resources across teams
Player-Coach	Builds while also developing people. Unlike traditional managers, their job is not information routing. Status meetings and alignment sessions are replaced by the World Model

The permanent middle management layer is eliminated.

What Startup CTOs Should Take Away

Block is a multi-billion dollar company with network effects from two platforms and transaction data that compounds daily. Their model will not translate directly to a 15-person startup. But several ideas from this essay are immediately applicable.

1. Audit Your Managers’ Time and Offload Information Routing

Most startups add their first managers around the 10-person mark. But look closely at what those managers actually do: translating leadership direction for the team, reporting team progress upward, coordinating with other teams.

The essay makes the case that this information-routing function is replaceable by AI. The first step for any CTO is to audit how your managers spend their time. If more than 60% goes to information routing, there is meaningful room for tooling to take over.

2. Make Work Machine-Readable

A critical prerequisite for Block’s model is remote-first operations. When work happens remotely, decisions, discussions, designs, plans, and progress all become text-based artifacts that AI systems can process.

Our organization operates fully remote, which means we already meet this prerequisite. Slack threads, GitHub discussions, and document-based decision-making naturally produce the text artifacts that Block’s World Model consumes. For fully remote teams, Block’s model is not a distant vision but rather a matter of better leveraging the artifacts you are already generating.

That said, having text artifacts is necessary but not sufficient. They need to be structured in ways that AI can effectively reference.

Concrete steps:

• Record all architectural decisions as ADRs (Architecture Decision Records)
• Transcribe Slack and Discord threads into documents with decision summaries
• Record meetings and generate transcripts using Claude or Whisper

3. Adopt the DRI Model

Block’s DRI (Directly Responsible Individual) model is directly adoptable at any scale.

Traditional “team leads” and “project managers” are permanent roles. DRIs are time-bounded. “Solve this problem in 90 days” with a clear scope and full authority to pull resources across teams for that period.

For resource-constrained startups, appointing a DRI per critical initiative is often more practical than maintaining permanent project management roles. The DRI model naturally fits organizations that cannot afford dedicated PMs.

4. Let Failure Signals Generate Your Roadmap

The Intelligence Layer’s design philosophy offers the essay’s most transferable insight. When the system cannot compose a solution for a user need, that gap is the product roadmap.

Most startups rely on PMs or founders building hypothesis-driven roadmaps. But if you instrument your product to detect cases where the current system cannot serve users, you get a more accurate signal for what to build next.

As the essay states:

Money is the most honest signal in the world.

Transaction and behavioral data reveal system limitations more accurately than PM intuition. Startup CTOs should design systems that automatically surface “cases the current product cannot handle” from logs and metrics.

5. Ask “What Does AI Reveal?” Not “What Does AI Replace?”

The essay closes with its most important question:

What can AI coordinate that hierarchy currently coordinates in your organization? If the answer is nothing, AI is just a cost optimization story. If the answer is deep, AI doesn’t augment your company. It reveals what your company actually is.

“AI makes us write code faster” and “AI lets us cut headcount” are cost optimization framings. What Block proposes is treating AI as something that changes the fundamental assumptions of organizational design.

The question for startup CTOs is not “what gets more efficient with AI?” but rather “if AI handles information routing, what should our organization actually look like?”

Practical First Steps

Block itself acknowledges this transition is early and imperfect:

Block is in the early stages of this transition. It will be a difficult one, and parts of it will likely break before they work.

No startup needs to adopt this model wholesale. But three things can start today:

1. Audit your managers’ time: Measure the percentage spent on information routing versus direct contribution
2. Make decisions machine-readable: Adopt ADRs and ensure all technical decisions are recorded in structured, searchable formats
3. Try a DRI: For your next critical initiative, appoint a time-bounded DRI with full cross-team authority, and compare the results to traditional project management

The future of organizational design is not the complete abolition of hierarchy. It is identifying which functions that hierarchy has been performing can be transferred to AI, and migrating them incrementally. Block’s “From Hierarchy to Intelligence” is one of the most systematic articulations of that direction to date.

That’s all from someone rethinking organizational design for the AI era. From the gemba.

References

• From Hierarchy to Intelligence (Block, Inc.)
• From Hierarchy to Intelligence (Sequoia Capital)
• Jack Dorsey’s post on X