Mini-AGI Companies: Replacing Hierarchy with AI Intelligence
Jack Dorsey argues that companies should be restructured as mini artificial general intelligences, replacing traditional hierarchies with AI-powered decision systems. By centralizing all organizational artifacts (Slack messages, code, documents) into an intelligence layer, companies can flatten structures, reduce layers, and let employees query real-time company data. Block laid off 40% of staff to rebuild this way, reducing organizational depth from five layers to two-three, and normalizing three core roles: individual contributors (ICs), directly responsible individuals (DRIs), and player-coaches. The shift treats AI as core infrastructure, not a productivity add-on.
The Problem with Traditional Hierarchy
Hierarchy Exists to Manage Information Flow
Traditional organizational hierarchies were designed to relay information across large groups of people at human scale. However, in remote-first companies where every action creates digital artifacts (Slack messages, emails, code, documents), this human relay system becomes inefficient and lossy.
Information Loss in Command Chains
Hierarchies rely on humans moving information up and down chains of command, introducing bias, politics, emotion, and agenda-driven filtering. This means leadership never sees the full picture of what the company is actually doing, failing at, or building.
The Mini-AGI Company Model
Treating Companies as Artificial General Intelligences
A company is already an intelligence—it processes information, makes decisions, and adapts. The innovation is structuring it efficiently by building a world model: an AI system trained on all organizational artifacts that can be queried by anyone in real time, making the company's state fully legible and enabling real-time decision-making.
Four-Layer Architecture for Modern Companies
Block's model consists of: (1) Capabilities—financial tools like card issuance and lending; (2) Interfaces—Square, Cash App, and other customer touchpoints; (3) Proactive Intelligence—AI that suggests actions based on transaction data and customer behavior; (4) World Models—deep understanding of customers and company state that enables real-time composition of features.
Customers Drive the Roadmap Through Queries
Instead of a fixed roadmap, customers interact with the company's intelligence and ask for features that don't exist. These queries reveal gaps in capabilities, which become the actual roadmap. The company then judges whether to build based on alignment with values and strategy.
Organizational Restructuring at Block
Three Core Roles Replace Traditional Hierarchy
Block is normalizing to three roles: (1) Individual Contributors (ICs)—builders and operators augmented by AI agents; (2) Directly Responsible Individuals (DRIs)—owners of customer outcomes and strategy; (3) Player-Coaches—humans who build capability in others by doing the work alongside them, not commanding.
Flattening Organizational Depth
Block reduced maximum organizational depth from five layers to a target of two-three layers, with the ideal being all 6,000 employees reporting directly to the CEO through the intelligence layer. This is manageable because the AI system handles information flow, not humans.
The 40% Layoff: Rebuilding for AI-First Operations
In December 2024, Block laid off 40% of staff (roughly 2,400 people) to rebuild the company as an intelligence. The decision came from asking: What is the minimum number needed to keep services running 100%, meet regulatory requirements, and grow while rebuilding as an AI-first company? The process took three weeks and was driven by principle, not panic.
CEO Role Transforms from Command to Alignment
The CEO's job shifts from ensuring principles, making decisions, and raising execution bars to constantly aligning the company's intelligence toward the right outcomes. Humans sit at the edge of the system, steering the AI-powered intelligence toward customer and strategic goals rather than making day-to-day decisions.
Board Building and Governance
Treat Investor Board Members as Recruiting Decisions
The first board member from an investor should be treated as a hire you can never fire (they can fire you). Optimize for the person who will uplevel conversation and execution, not brand name. This relationship has far greater impact on company outcome than the financing terms.
Add Independent Board Members Early
Get a strong independent board member within one to two years, certainly by product-market fit. Independent directors offer a different dynamic than investor directors and can provide valuable mentorship without the inherent conflict of interest.
Build Boards with Care, Not in a Rush
Boards are often assembled too late, especially before IPO, resulting in members with no context or chemistry. When tested by crises (short sellers, hostile investors, financing challenges), weak boards fail. Build boards intentionally over time with aligned values and proven dynamics.
CEO Challenges: Multi-Company Leadership and Second Acts
Being CEO of Two Public Companies Is an Anti-Pattern
Jack Dorsey's simultaneous leadership of Twitter and Block should not be a goal for others. The divided attention and conflicting priorities make it nearly impossible to execute at the highest level. Private companies might work differently, but public companies require full focus.
Delegating Too Much Creates Holding Company Dynamics
Jack's biggest regret was over-delegating at Block, creating multiple CEOs for different divisions (Square, Cash App). This fragmented the company into unrelated businesses growing at different rates instead of leveraging both sides of the financial network together.
Building Second Acts Requires Credibility Loss
Cash App and Square's lending business both faced internal and board resistance. Success required Jack to lose credibility with stakeholders while defending the new initiative, then earn it back through results. Founders must be comfortable with temporary credibility loss for strategic bets.
The DGAF Principle: High Conviction with Humility
Jack's strength and weakness is not giving a flock—he sticks with unpopular convictions. But this works only when paired with genuine learning from feedback and willingness to be wrong. The key is maintaining conviction while remaining open to course correction.
Modern CEO Skills and Mindset
Every Encounter Is a Mentor
Rather than relying on a single CEO coach, Jack shifted to treating every person, problem, and encounter as a teaching moment. This requires deciding to learn from each interaction and writing down the lesson. Biggest regrets come from deciding not to learn something.
Meditation Sharpens Focus and Observation
Jack practices intensive meditation retreats (10 days, 4:30 AM to 9 PM focus on breath sensation, then body scanning). This trains the mind to focus sharply, observe without emotional reaction, and recognize impermanence. Applied to leadership, it diminishes reactive instinct and enables deliberate choice.
Timeless CEO Qualities: ALE
Roelof Botha's framework for enduring CEO traits: Authenticity (being genuine, not pretentious), Logic (predictability, rationality), and Empathy (caring deeply about team and business). These remain constant despite AI transformation.
Reprogramming Assumptions Constantly
The new critical skill is the ability to question your own requirements, assumptions, and past decisions. As tools accelerate change, the ability to break free from momentum and sacred cows becomes more valuable than ever.
AI as Input, Not Output
Most people treat AI outputs as final decisions. Jack treats them as input to create better output. This requires taste, point of view, and judgment—the ability to discern signal from noise and decide what's relevant and opinionated.
Practical Implementation and Transformation
Meetings Evolve from Slides to Prototypes
Two months into Block's transformation, meetings shifted from presentations and Google Docs to employees bringing working prototypes with real or simulated data. This enables real-time conversation and modification, allowing teams to explore breadth and depth simultaneously.
Exploration Becomes Faster and Cheaper
With AI tools, exploring multiple paths costs near-zero. Hardware prototypes that took weeks now take hours. This enables more parallel exploration, but focus remains critical on getting the last 20% right—the part requiring creativity, taste, and judgment.
The Pace of Change Demands Speed
The speed at which Block executed this transformation (three weeks from concept to announcement) reflects the urgency of AI-driven disruption. Dithering for months would have meant falling behind. Moving fast with principle beats moving slow with caution.
Distribution and Attention Become New Differentiators
As AI tools commoditize building speed, distribution and attention become the new moats. With so much noise, the ability to reach customers and stand out becomes more critical than the ability to build features.
The Existential Question: What Is a Company?
Existential Dread and Opportunity Coexist
Jack has experienced simultaneous existential dread and optimism about what companies will become. The question 'what is even a company going forward' is the central challenge. The mini-AGI model is the only durable structure he can imagine lasting.
AI Is a Structural Shift, Not a Productivity Layer
Most companies view AI as a co-pilot that makes everyone 10x more productive. Jack argues this is wrong—it's a structural transformation requiring complete rethinking of how organizations work, not an add-on to existing hierarchies.
Customers Will Expect Infinite Customization
As AI enables real-time composition of features, customers will expect to ask for futures that don't exist on a roadmap and have them served immediately. This removes the roadmap as a limiting factor and forces companies to think in terms of capabilities and composition.
Money as the Honest Signal
Transaction data is the most honest signal of what customers actually want and need. Unlike surveys or interviews, money doesn't lie—when a transaction occurs, it reveals truth about a person's or business's life and priorities.
Notable quotes
I don't think this is a productivity thing. I think it's a structural thing that needs to shift. — Jack Dorsey
Money is the most honest signal in the world. When a transaction occurs, that really tells the truth. — Jack Dorsey
Every single person I talk with, every encounter I have, that's my mentor. — Jack Dorsey
Action items
- Audit all digital artifacts your company creates daily (Slack, email, code, documents, meetings) and assess what intelligence could be built on top of them.
- Map your current organizational depth—count layers from CEO to individual contributor—and set a target to reduce by 30-50%.
- Define the three roles for your company: individual contributors (with AI augmentation), directly responsible individuals (DRIs), and player-coaches (capability builders).
- Eliminate or consolidate middle management layers by building a query-able company intelligence system that makes decisions transparent and legible.
- Start treating AI outputs as input to your decision-making, not final output. Build taste and judgment into how you evaluate AI suggestions.
- Shift from fixed roadmaps to capability-based architecture: define core capabilities, interfaces, proactive intelligence, and world models.
- For board building: treat investor board members as recruiting decisions, add independent directors early, and build boards intentionally with aligned values.
- Practice daily learning: write down one lesson from every encounter, conversation, and problem you face.
- Question your sacred cows and past decisions regularly. Build a culture where assumptions are constantly reprogrammed.
- Move fast on structural changes. Dithering for months means falling behind in an AI-driven world.