On 31 March 2026, Jack Dorsey and Roelof Botha of Sequoia Capital published a piece called “From Hierarchy to Intelligence”. If you haven’t read it, you should — it is one of the most significant things a sitting tech CEO has put in public in years. Not because it is radical theory, but because it describes something that is already happening at Block, and that will eventually happen everywhere.
The argument is simple: the organisational hierarchy that has governed every large company since the Roman Army exists for one reason — to route information. AI can now do that better than humans. So the hierarchy, as we know it, becomes optional.
Most leaders reading this will think: “Interesting. Not for us yet.” And they’re right — most UK businesses are nowhere near ready for what Dorsey describes. But here’s where I’d push back: not being ready to implement something is not the same as not needing to think about it. The businesses that will win over the next five years are already thinking across three levels of AI transformation simultaneously. Most are only thinking about one.
The Short Answer
The Dorsey piece describes the endgame of AI adoption: a company organised as an intelligence rather than a hierarchy, where AI replaces the information-routing function of management and humans focus on judgement, craft, and edge-case decisions. Most UK mid-market businesses are at the beginning of their AI journey — using AI for productivity and cost reduction. That is necessary, but it is not sufficient. At Fifty One Degrees, we work with founders and senior leaders across UK businesses and the pattern is consistent: the companies building a durable competitive advantage are not just automating tasks. They are thinking simultaneously across three layers — productivity, revenue intelligence, and organisational architecture — even if the third layer is embryonic. The gap between businesses operating at one layer and businesses operating at all three will compound. In three to five years, it will be very difficult to close.
What Dorsey Actually Said — And Why It Matters
The Sequoia piece opens with a history of organisational design that is worth reading in full. Dorsey traces the modern corporate hierarchy from the Roman Army through the Prussian General Staff to the American railroads and Frederick Taylor’s scientific management. The conclusion: every organisational structure ever devised is essentially an information routing protocol, built around a simple human constraint — a leader can effectively manage somewhere between three and eight people.
Add more people, add more layers. Add more layers, slow down information flow. Two thousand years of organisational innovation has been an attempt to work around this tradeoff without ever truly breaking it.
Dorsey argues that AI removes this constraint entirely. Block is building what he calls a world model of its own operations — a continuously updated, machine-maintained picture of what is being built, what is blocked, and where resources are allocated. In a traditional company, that picture lives in managers’ heads and travels up and down a chain of command. At Block, AI carries it instead.
On top of that world model sits an intelligence layer — a system that composes existing capabilities into solutions for specific customers at specific moments, without a product manager hypothesising what to build. Below it are atomic building blocks — payments, lending, card issuance — combined and recombined by the intelligence layer as needed, with no user interfaces of their own.
The org structure that follows from this collapses to three roles: individual contributors who build and operate the system; directly responsible individuals who own cross-cutting outcomes for defined periods; and player-coaches who combine building with developing people. No permanent middle management layer. The system coordinates what humans used to.
This is not science fiction. Block is doing it now.
Why Most UK Businesses Are Only Playing Layer 1
The honest picture of AI adoption in UK mid-market businesses right now is this: most are at Layer 1. They are buying licences, running training sessions, automating individual tasks, and seeing genuine productivity gains. That is real and it is valuable. According to the BCC’s Powering Productivity report published in March 2026, 54% of UK SMEs are now actively using AI — up from 35% the previous year. But adoption is shallow. Companies are adding AI to existing processes rather than rethinking the processes themselves.
This is what I’d call the Productivity Ceiling. You can get meaningful efficiency gains from Layer 1 — time saved, cost reduced, output increased. But the ceiling is relatively low, because you are still running the same underlying operating model. You’ve given everyone a faster car. You haven’t changed the road.
The deeper problem is that Layer 1 thinking does not prepare you for what comes next. If the only question your leadership team is asking about AI is “how do we use it to do what we already do, faster?” — you are going to be surprised by competitors who asked a different question.
We have written previously about getting your entire team using AI — the training, culture, tools, and operating practices that drive genuine adoption. That is the essential first step. But it is the first step, not the destination.
The Three-Layer AI Play
At Fifty One Degrees, we think about AI value creation across three distinct layers. These are not sequential stages — you do not complete Layer 1 before starting Layer 2. They run in parallel, at different speeds and with different levels of maturity. The businesses building structural advantage are operating at all three simultaneously.
- Layer 1 — Productivity: AI as a tool. Copilots, automation, task-level efficiency. Reducing cost, increasing output, freeing up time. This is where most businesses are today, and where every business should start. The question is whether it is where they stop.
- Layer 2 — Revenue Intelligence: AI as a growth engine. Predictive models that move a business from hindsight to foresight — customer churn prediction, demand forecasting, lead scoring, dynamic pricing, lifetime value modelling. This is where AI stops being a productivity tool and starts being a commercial weapon. The businesses operating at Layer 2 are not just working smarter. They are making better decisions, faster, with information their competitors do not have.
- Layer 3 — Organisational Intelligence: AI as the operating system. This is what Dorsey describes — replacing the information-routing function of hierarchy with a machine-maintained model of the business. It is the structural endgame. Most UK businesses will not reach full Layer 3 implementation in the near term. But the leaders who begin thinking about it now — who start asking “what does our business understand that is genuinely hard to understand, and is that understanding getting deeper every day?” — will be the ones positioned to get there.
The core insight of The Three-Layer AI Play is not that the layers are different. It is that they are interdependent. Layer 2 requires the data infrastructure built at Layer 1. Layer 3 requires the intelligence built at Layer 2. You cannot skip, but you can run them in parallel — and running them in parallel is what separates the businesses building a compounding advantage from those chasing a static efficiency gain.
What Organisational Intelligence Actually Looks Like at a Human Scale
Dorsey’s piece is written at the scale of Block — a global fintech with thousands of employees and millions of transactions. It is tempting for the leader of a £50m UK business to read it and conclude it does not apply to them. That would be a mistake.
The principles scale down. Every business has a version of the world model problem: information that should be visible across the organisation but instead lives in managers’ heads, email threads, and spreadsheets. Every business has a version of the intelligence layer problem: decisions that are delayed because the right data is not in front of the right person at the right moment. The technology to address these problems at mid-market scale already exists.
At Fifty One Degrees, we operate with a delivery-to-leadership ratio of 10:1 — ten people doing the work for every one coordinating it. That is an unusually lean operating model for a professional services firm. It is achievable because we have automated many of the coordination and information-routing functions that would otherwise require additional management layers. The result is faster decisions, less overhead, and more of our team’s time spent on client work that creates direct value. It is not the full Dorsey model. But it is the same principle, applied at our scale.
The question for any leader is not “can we become Block?” The question is: which of your management layers exist primarily to route information, and what would it take to give that job to a system instead?
| Traditional Model | Intelligence-Led Model |
|---|---|
| Information travels up and down a chain of command | AI maintains a continuously updated picture of operations |
| Middle management coordinates and relays context | Individual contributors act with full context from the system |
| Roadmap set by product managers hypothesising demand | Customer reality and unmet capability gaps generate the backlog directly |
| Hierarchy adds layers as headcount grows | Span of control expands without adding coordination overhead |
| Speed limited by human information flow | Speed limited only by capability and judgement |
What Leaders Should Do Right Now
The gap between where most UK businesses are and where Dorsey is pointing is real. Closing it in full will take years. But there are practical things a leader can do now — this quarter — that start building in the right direction across all three layers simultaneously.
Audit your management layers for information routing
Ask honestly: which of your management roles exist primarily to gather information from below, synthesise it, and pass it upward? Those roles are the ones AI will displace first. That is not an argument for making people redundant today — it is an argument for understanding where your organisational dependencies are, and starting to build alternatives.
Identify your proprietary signal
Dorsey’s argument hinges on Block’s access to both sides of millions of transactions — a data signal that compounds in value over time. Every business has some version of this: data that is genuinely hard for a competitor to replicate. What is yours? Customer behaviour, operational performance, financial patterns? If you do not know the answer, that is the first problem to solve.
Start Layer 2 while you are still running Layer 1
Most businesses treat predictive modelling and revenue intelligence as something they will get to eventually. The businesses that treat it as a parallel workstream — running alongside the productivity and automation work — build a two-to-three year head start. The data models you build today become the intelligence layer of tomorrow.
Read the Sequoia piece with your leadership team
Not as an exercise in abstract strategy, but as a practical provocation: which of the four elements Dorsey describes — capabilities, world model, intelligence layer, interfaces — do you have the beginnings of already? Which are entirely absent? The answers will tell you a great deal about where your AI strategy actually is, versus where you think it is.
Work with people who are already building this
The fastest way to move across all three layers is to embed people who have already done it. Not consultants who advise from the outside — practitioners who sit inside your team and build alongside you. Fifty One Degrees was founded on that model specifically because we believe it is the only one that produces lasting change rather than a well-crafted slide deck.
Frequently Asked Questions About AI and Organisational Design
What is Jack Dorsey’s “From Hierarchy to Intelligence” article about?
Published on 31 March 2026 by Sequoia Capital, it argues that organisational hierarchy exists solely to route information across management layers — and that AI can now perform that function better than humans. Dorsey describes how Block is rebuilding its operating model around a world model and an intelligence layer, replacing traditional management with three roles: individual contributors, directly responsible individuals, and player-coaches.
Does the Dorsey model apply to mid-market UK businesses, or only large tech companies?
The principles apply at any scale. The underlying problems — information routing, coordination overhead, decision latency — exist in every business above a few dozen people. At Fifty One Degrees, we are already applying these principles with UK mid-market clients, building the infrastructure that makes Layer 2 and Layer 3 thinking practical at that scale.
What is the difference between using AI for productivity and using AI to replace hierarchy?
Productivity-level AI makes existing processes faster and cheaper — automating tasks, drafting content, summarising information. Replacing hierarchy means redesigning how your organisation coordinates and routes information — using AI to maintain the contextual awareness that managers currently provide. The difference is the depth of change: productivity improves what you do; organisational intelligence changes how your business works.
How long will it take UK businesses to reach the organisational intelligence model Dorsey describes?
For most UK mid-market businesses, full Layer 3 implementation is three to seven years away — assuming they start building the data infrastructure now. Businesses that begin today and run all three layers in parallel will compress that timeline significantly. Those that stay at Layer 1 will fall behind competitors who did not.
Should I be worried about AI replacing my management team?
Not imminently, but directionally — yes, for roles whose primary function is information routing and coordination. Roles involving judgement, client relationships, and cultural leadership are far more durable. The more useful question is: are your managers spending most of their time on information routing, or on the things AI cannot do? Shifting that balance is something every business can start on today.
What does a “world model” look like for a £50m UK business?
The equivalent might start with a live operational dashboard fed by automated data pipelines, AI that can answer questions about pipeline and capacity without a manager preparing a report, and predictive models that flag problems before they surface in a weekly meeting. Fifty One Degrees helps UK businesses build this infrastructure as part of a structured AI strategy engagement.
How do I know which layer my business is currently operating at?
Layer 1: AI tools for task-level productivity. Layer 2: predictive models informing commercial decisions — churn, demand, lead quality, pricing. Layer 3: AI involved in how your business coordinates and routes information without requiring a human relay. The Three-Layer AI Play framework, used by Fifty One Degrees when scoping AI strategy engagements, helps leadership teams map where they are and what building all three layers in parallel looks like in practice.
The Gap Is Opening Now
Dorsey’s piece is not a prediction about the distant future. It is a report from the present — written by a CEO who is already doing it. The businesses that read it, take it seriously, and start asking the right questions will look back in five years and identify this moment as when the gap between them and their competitors began to open.
The honest truth is that most UK businesses are at Layer 1. Getting Layer 1 right is the essential foundation — if you have not yet done that work, start there. But do not let Layer 1 become the ceiling. The leaders who will define the next five years are already thinking about all three layers at once.
Want to understand where your business sits across The Three-Layer AI Play? Talk to us at Fifty One Degrees — we’ll help you map what building all three layers in parallel looks like for your specific context.
