Jack Dorsey's Mini-AGI: What Happens When a Company Becomes the Intelligence

Jack Dorsey is doing something at Block that most companies won't attempt for another five years. He's not just adding AI tools to an existing org chart. He's rebuilding the org chart around AI as the connective tissue. The result is what he calls a "mini-AGI" — and whether you agree with the approach or not, the underlying ideas are worth understanding.

From hierarchy to three roles

Traditional corporate structures are layers of management. Directors report to VPs who report to SVPs who report to the C-suite. Information flows up slowly, decisions flow down even slower, and by the time anything reaches the people doing the work, context has been lost at every handoff.

Dorsey's answer is radical simplification. Block is moving to three roles:

Individual Contributor (IC) — the person doing the work. Writing code, designing products, making the thing. No manager sitting between them and the outcome.

Directly Responsible Individual (DRI) — the person who owns the outcome. Not a manager in the traditional sense. A DRI doesn't approve your pull request or schedule your one-on-one. They own a result, and they're accountable when it doesn't happen.

Player-Coach — the person who builds the team's capacity while still contributing directly. Not a pure manager. Not a pure IC. Someone who can do the work and help others get better at doing it.

That's it. Three roles. The entire management layer between "person doing the work" and "person accountable for the result" collapses into a system where ownership is explicit and coaching is embedded.

The company as intelligence layer

Here's where it gets interesting. Dorsey's insight isn't just organizational — it's architectural. Every company already produces a massive stream of digital artifacts: Slack messages, code commits, design documents, meeting transcripts, dashboards, support tickets. In most organizations, this information is siloed. The engineering team doesn't see what support is hearing. Product doesn't know what's actually shipping versus what's planned.

Dorsey's concept of the "mini-AGI" is what happens when you layer AI across all of these artifacts simultaneously. Not a chatbot that answers questions. An intelligence layer that can surface patterns, connect decisions to outcomes, and make the company's collective knowledge accessible to anyone who needs it.

Think about what this means in practice. A DRI working on payment processing doesn't need to schedule meetings with six teams to understand the current state. The intelligence layer has already synthesized the code changes, the support tickets, the design docs, and the metrics. The DRI can ask a question and get an answer grounded in what's actually happening — not what someone remembers from last Tuesday's standup.

This is the same principle behind agent architectures: give the system access to the right context, and it can reason about that context in ways that individual humans — limited by attention, memory, and organizational boundaries — simply cannot.

Founding moments don't just happen once

One of the more nuanced ideas Dorsey raises is the concept of recurring "founding moments." Most people think of a company's founding as a single event — the garage, the first commit, the napkin sketch. But Dorsey argues that companies experience multiple founding moments throughout their lifecycle.

A new strategic direction is a founding moment. An internal team discovering a capability no one planned for is a founding moment. A technology shift that changes what's possible — like the current wave of AI — is a founding moment.

The problem with traditional hierarchies is that they're optimized for execution, not for recognizing founding moments when they happen. Information gets filtered through layers of management, and by the time a breakthrough idea reaches someone with the authority to act on it, it's been sanded down into a quarterly roadmap item.

The mini-AGI model is designed to make founding moments visible. When the intelligence layer can surface unexpected patterns — a support trend that suggests a new product, a code change that enables a capability no one anticipated — the company can respond like a startup, even at scale.

The skills that survive

If AI handles information synthesis and organizational context, what's left for humans? Dorsey's framework maps specific skills to each of the three roles:

ICs need judgment, taste, and creativity. The intelligence layer can tell you what's happening. It can even suggest what to do. But deciding what's worth building — what's elegant versus what's merely functional, what serves users versus what serves metrics — that requires human judgment that AI can inform but not replace.

DRIs need ownership and accountability. This is the hardest skill to automate because it's fundamentally about accepting consequences. When a DRI says "this is my responsibility," they're making a commitment that no AI system can make on their behalf. The intelligence layer makes accountability more visible — you can't hide behind information asymmetry — but the willingness to own outcomes remains distinctly human.

Player-Coaches need the ability to build capacity in others. Teaching someone to be better at their craft isn't just knowledge transfer. It's understanding where someone is stuck, what mental model they're missing, and how to create the conditions for them to figure it out themselves. AI can augment coaching with data — showing patterns in someone's work, identifying skill gaps — but the coaching relationship itself is human.

What this means for everyone else

You don't have to be building a payments company to take something from Dorsey's playbook. The core insight is structural: most organizations have too many roles dedicated to moving information between people, and AI makes many of those roles unnecessary.

The question isn't whether your company should adopt Block's exact three-role model. It's whether you're honest about which layers of your organization exist to do work and which exist to manage the flow of information. Because the information management layer is exactly what AI is about to absorb.

The companies that figure this out early won't just be more efficient. They'll be more responsive, more transparent, and better at recognizing the founding moments hiding in their own data.

Credit to Jack Dorsey for articulating this vision publicly. Whether Block's experiment succeeds or not, the ideas deserve serious consideration from anyone thinking about what organizations look like in an AI-native world.