By Thomas Thomison
If Holacracy, Sociocracy, Reinventing Organizations, Teal, or the wider self-management movement changed how you understand authority and purpose, I have good news: you are standing at the edge of a future that makes your work a prerequisite.
You have been training for it.
Not because self-management anticipated every capability now arriving through artificial intelligence. It did not. And not because adding a few agents to a circle structure magically produces the organization of the future. It will not.
You are unusually prepared because you have already had to unlearn several assumptions that most organizations still treat as natural law: that authority must live in a person; that a job and the human holding it are the same thing; that coordination requires a manager; that disagreement is friction; that planning is steering; and that organizational structure is something drawn once, announced, and gradually ignored.
Those assumptions were already expensive. In the Age of Intelligence, they become architectural liabilities.
When intelligence can do more than answer questions—when it can hold memory, draft and revise artifacts, invoke tools, coordinate across systems, participate in governance, and increasingly take consequential action—the central organizational question changes. Intelligence becomes a forcing function for change—and maybe it always has.
It is no longer merely: What can the AI do?
It becomes: Under what authority does it act?
That is a governance question. And it is one the self-management movement has been preparing to answer for decades.
My own path into this question began in 2007, when I co-founded HolacracyOne. We were not trying to prepare organizations for artificial Minds. We were trying to answer a more immediate question: how can rules and purpose, rather than a person, hold final authority—and how can the structure of work evolve from the tensions sensed by anyone doing it?
That line continued through encode.org and the For-Purpose Enterprise, where we worked to differentiate capital, labor, operations, and legitimate governance, and continues now with PowerShift® Intelligence. In retrospect, these look less like separate experiments than successive attempts to make authority explicit enough to travel across time, scale, and substrate. The lineage from Holacracy to the Algorithmic Organization is not a pivot away from self-management. It is the next scale of the same inquiry.
I define an Algorithmic Organization as an organization whose roles, rules, policies, and learning loops are explicit enough to be run by human and artificial agents alike.
That is not the same as a company using AI agents. Every conventional company will soon use agents, just as every company eventually used email. Tool adoption does not produce a new organizational form.
An Algorithmic Organization is different at the level of its operating logic. It can coordinate effort, allocate resources, and align incentives across a field of human and artificial intelligence because authority is carried by explicit structure rather than inferred from status, proximity, or personality. It is intelligence at the coordination layer.
This distinction matters because intelligence is not inherently aligned. I define intelligence as the rate at which a system metabolizes the tension it encounters into purposeful change. A bacterium can do this. A human can do this. A team, an AI agent, and—uncomfortably—a cancer cell can do this.
The cancer cell is material to the definition. It is highly capable at pursuing a local purpose that the larger whole cannot legitimately endorse.
So intelligence is not enough. The design target is governed intelligence: intelligence in service of a purpose legitimate at its own scale.
The word governed does real work. It means we do not merely hope an intelligent system remains aligned. We locate its purpose, authority, constraints, escalation paths, and correction loops in a structure the whole can inspect and change.
Culture is not an access-control system. A vibe is not an authority model, however excellent the offsite.
This is where self-management practitioners have an extraordinary head start.
Already fluent in self-management? The frontier begins at The inheritance is real. It is not sufficient.
Most conventional organizations are difficult enough for humans to interpret. Job titles mix identity, status, skill, authority, and social position into a single object. Decision rights often depend less on written policy than on proximity to power. Exceptions live in private memory. The org chart says one thing; the actual network of permission says another.
Humans survive this through years of tacit learning. We read tone, infer hidden boundaries, remember who must be consulted, and discover which rules are aspirational. Much of what we call “organizational experience” is skilled navigation of institutional ambiguity.
Artificial agents do not remove that ambiguity. They accelerate whatever is already there.
Place an agent inside an unclear organization and it will become either a very powerful intern—waiting for permission at every consequential threshold—or an invisible executive, exercising authority no one remembers granting. Neither condition is an intelligence strategy. Both are governance failures with excellent throughput.
This is why adding AI to management is not the same as building an Algorithmic Organization. Automating ambiguity simply produces ambiguity at machine speed. A prompt is not a constitution. Permission is not authority. A capable model is not, by capability alone, a legitimate organizational actor.
Self-management practitioners begin closer to the real problem because we have already spent years making the invisible structure of work visible.
Holacracy, Sociocracy, Teal, and the wider self-management movement are not interchangeable. Holacracy and Sociocracy offer explicit governance grammars; Teal names a developmental pattern of evolutionary purpose, self-management, and wholeness that Frédéric Laloux helped a much broader audience recognize. Their differences matter. Their shared inheritance is a set of practiced instincts that conventional management still struggles to make operational.
The work can be separated from the worker. Roles describe functional work, purpose, authority, and accountabilities without reducing the person filling them to a job. That makes work portable across humans and Minds while preserving the distinction between role and soul.
Authority can live in structure. Constitutions, circles, roles, domains, policies, and governance records allow legitimate authority to move beyond personality and proximity. Artificial agents need this structural answer to what they may do, for what purpose, within which bounds, and where human judgment remains required. This is the difference between autonomy and drift.
Tensions, objections, and drivers are information. These practices teach the system to metabolize discrepancy. My definition makes the connection explicit:
Intelligence is the rate at which a system metabolizes the tension it encounters into purposeful change.
Capability alone remains neutral; a cancer cell is highly capable at pursuing a local purpose the organism cannot legitimately endorse. So the second definition carries the governance burden:
Governed intelligence is intelligence in service of a purpose legitimate at its own scale.
Governance and operations are different kinds of work. An agent may operate broadly within a role while lacking authority to expand that role, rewrite its constraints, enlarge its budget, or change the policy governing its actions. Operational agency can be fast. Structural change remains explicit, reviewable, and legitimate.
Purpose and boundaries create local freedom. Purpose acts as a relevance filter; boundaries give intelligence somewhere coherent to act. Rules in service of freedom. The paradox becomes practical once you have lived it.
Clarity is an operating condition. Roles, policies, domains, decisions, and tensions must be visible enough for distributed actors to coordinate. In an Algorithmic Organization, that hard-won transparency becomes machine legibility. An authority that exists only in someone's head cannot reliably govern a Mind.
Good objections improve the whole. Mixed human–AI systems need agents able to surface tensions, reveal inconsistencies, and challenge assumptions without confusing computational fluency with moral standing or final authority. Standing is a governance grant, not a metaphysical verdict.
Self-management has been teaching us to release distributed agency while retaining coherence. That is no longer a specialist concern. It is the alignment problem inside the organization.
There is a tempting but unhelpful conclusion available here: We already solved this.
We did not.
Holacracy is not an Algorithmic Organization. Sociocracy is not an Algorithmic Organization. A self-managing culture does not become agent-ready by opening an API account and assigning a chatbot to a circle. If we down-translate the new frontier into our familiar frameworks, we will lose the frontier while congratulating ourselves on recognizing it.
The inherited governance kernel must be extended.
An intelligence-bearing organization needs explicit ways to establish an agent’s identity, purpose, capabilities, constraints, and governance standing. It needs memory that can be inspected and corrected; receipts for consequential handoffs; budgets for money, compute, tokens, and attention; escalation paths; versioned policies; proof standards; and hard prohibitions against unauthorized self-expansion. It needs legal bridges between constitutional governance and the entities that hold assets and liability. It needs human judgment at the points where legitimacy cannot be delegated to pattern recognition.
Most of all, it needs a more precise human role.
I call that role the Orchestrator: a human who convenes, stewards, and calibrates a field of intelligence in service of purpose.
The old manager pushed work through people. The Orchestrator holds purpose, senses tensions, stewards authority, allocates scarce attention and resources, calibrates Minds, and keeps the system honest. This is not “managing agents” with newer nouns. It is a different relationship to intelligence.
And here again, self-management practitioners have an advantage. You have already practiced the underlying moves:
These are not the whole practice of orchestration. They are unusually good soil for it.
For years, self-management has often been treated as an organizational specialty, a niche within a niche: compelling to a subset of founders, consultants, cooperatives, intentional communities, and teams willing to tolerate unusual meeting vocabulary.
I think that frame must now expand.
The movement has functioned as a distributed research program into post-managerial coordination. Its living laboratories have been imperfect and sometimes exhausting. Some implementations moved authority; others mostly changed the meeting format. Some produced real sovereignty; others concealed a founder hierarchy beneath more elegant language. Many of us have accumulated scar tissue along with insight.
Good. Scar tissue is data—provided the injury was survivable.
The next generation of organizations will need people fluent in authority without dominance, structure without centralization, autonomy without drift, and continuous change without perpetual reorganization. It will need practitioners who understand that sensing a tension is not insubordination, that an objection can be an act of service, and that purpose must do real decision work, not simply inspire.
That is this community.
The opportunity is not to deify or defend Holacracy, Sociocracy, or self-management as the final answer. It is to bring forward what the practices have taught us, allow those teachings to evolve, and help build the governance layer for intelligence-integrated organizations.
No victory lap is required. The future is generally unimpressed by those.
The Age of Intelligence will not reduce the importance of human sovereignty. It will make sovereignty non-delegable, non-negotiable.
As artificial intelligence gains memory, tools, continuity, and operational agency, humans will be asked to do less of the work that once made us feel indispensable. Our contribution will move upstream: holding purpose, making legitimacy judgments, designing authority, attending to consequences, and deciding what kinds of systems deserve to exist.
That is not a diminishment of the human role. It is a clarification.
Those who have practiced self-management are unusually prepared for it—not because we have escaped hierarchy, ego, ambiguity, or power, but because we have learned to see them as design material for what's next. We have learned that empowerment is systemic, that autonomy needs structure, and that purpose is only real when it can govern a choice.
The Algorithmic Organization will require new runtimes, memory systems, legal forms, agent protocols, and technical safeguards. But beneath all of them sits an older question: can intelligence be given enough freedom to act while remaining in legitimate service to the whole?
Holacracy and Sociocracy have been working on that question all along.
So has everyone who has tried, seriously and sometimes painfully, to move authority out of the person and into a transparent practice.
The Age of Intelligence did not make that work quaint. It revealed what it was for: a prerequisite for what comes next.
You’ve been training for this all along. It is time to step forward and model what is now possible.
If you are already practicing Holacracy or Sociocracy, you have already done the hard part: you have learned to map the organization functionally and move authority into explicit practice. Start with one Circle—with or without Sub-Circles—and pilot governed agents as role-fillers. Give them a defined purpose, accountabilities, domains, capability envelope, constraints, and escalation paths. Keep human judgment where legitimacy requires it. Keep receipts. Let tensions update the structure.
If you are working in a conventional organization, start with a bounded pilot: one team, department, or consequential workflow. Map it functionally first—purpose, roles, authority, policies, and boundaries—then introduce governed agents inside that structure. Draw on experienced Holacracy or Sociocracy practitioners if useful, or build the map yourself. The point is not to automate the org chart. It is to make authority explicit enough for intelligence to act without drifting.
However you enter, we are here to support the journey. Go free or go fast with PowerShift® Intelligence: adopt the open constitutional layer yourself, or work with us to form and accelerate an Algorithmic Organization.
Tension is fuel. Purpose sets the bounds.
# You’ve Been Training for This ## Why Holacracy, Sociocracy, and the wider self-management movement are unusually prepared to lead the Algorithmic Organization *By Thomas Thomison* If Holacracy, Sociocracy, *Reinventing Organizations*, Teal, or the wider self-management movement changed how you understand authority and purpose, I have good news: you are standing at the edge of a future that makes your work a prerequisite. You have been training for it. Not because self-management anticipated every capability now arriving through artificial intelligence. It did not. And not because adding a few agents to a circle structure magically produces the organization of the future. It will not. You are unusually prepared because you have already had to unlearn several assumptions that most organizations still treat as natural law: that authority must live in a person; that a job and the human holding it are the same thing; that coordination requires a manager; that disagreement is friction; that planning is steering; and that organizational structure is something drawn once, announced, and gradually ignored. Those assumptions were already expensive. In the Age of Intelligence, they become architectural liabilities. When intelligence can do more than answer questions—when it can hold memory, draft and revise artifacts, invoke tools, coordinate across systems, participate in governance, and increasingly take consequential action—the central organizational question changes. Intelligence becomes a forcing function for change—and maybe it always has. It is no longer merely: **What can the AI do?** It becomes: **Under what authority does it act?** That is a governance question. And it is one the self-management movement has been preparing to answer for decades. My own path into this question began in 2007, when I co-founded HolacracyOne. We were not trying to prepare organizations for artificial Minds. We were trying to answer a more immediate question: how can rules and purpose, rather than a person, hold final authority—and how can the structure of work evolve from the tensions sensed by anyone doing it? That line continued through encode.org and the For-Purpose Enterprise, where we worked to differentiate capital, labor, operations, and legitimate governance, and continues now with PowerShift® Intelligence. In retrospect, these look less like separate experiments than successive attempts to make authority explicit enough to travel across time, scale, and substrate. The [lineage from Holacracy to the Algorithmic Organization](https://surface.powershift.io/lineage/index.html) is not a pivot away from self-management. It is the next scale of the same inquiry. ## The future arrives through a question of authority I define an **Algorithmic Organization** as an organization whose roles, rules, policies, and learning loops are explicit enough to be run by human and artificial agents alike. That is not the same as a company using AI agents. Every conventional company will soon use agents, just as every company eventually used email. Tool adoption does not produce a new organizational form. An Algorithmic Organization is different at the level of its operating logic. It can coordinate effort, allocate resources, and align incentives across a field of human and artificial intelligence because authority is carried by explicit structure rather than inferred from status, proximity, or personality. It is intelligence at the coordination layer. This distinction matters because intelligence is not inherently aligned. I define intelligence as **the rate at which a system metabolizes the tension it encounters into purposeful change**. A bacterium can do this. A human can do this. A team, an AI agent, and—uncomfortably—a cancer cell can do this. The cancer cell is material to the definition. It is highly capable at pursuing a local purpose that the larger whole cannot legitimately endorse. So intelligence is not enough. The design target is **governed intelligence: intelligence in service of a purpose legitimate at its own scale**. The word *governed* does real work. It means we do not merely hope an intelligent system remains aligned. We locate its purpose, authority, constraints, escalation paths, and correction loops in a structure the whole can inspect and change. Culture is not an access-control system. A vibe is not an authority model, however excellent the offsite. This is where self-management practitioners have an extraordinary head start. > **Already fluent in self-management?** The frontier begins at [*The inheritance is real. It is not sufficient.*](#the-inheritance-is-real-it-is-not-sufficient) ## Conventional organizations carry semantic debt Most conventional organizations are difficult enough for humans to interpret. Job titles mix identity, status, skill, authority, and social position into a single object. Decision rights often depend less on written policy than on proximity to power. Exceptions live in private memory. The org chart says one thing; the actual network of permission says another. Humans survive this through years of tacit learning. We read tone, infer hidden boundaries, remember who must be consulted, and discover which rules are aspirational. Much of what we call “organizational experience” is skilled navigation of institutional ambiguity. Artificial agents do not remove that ambiguity. They accelerate whatever is already there. Place an agent inside an unclear organization and it will become either a very powerful intern—waiting for permission at every consequential threshold—or an invisible executive, exercising authority no one remembers granting. Neither condition is an intelligence strategy. Both are governance failures with excellent throughput. This is why adding AI to management is not the same as building an Algorithmic Organization. Automating ambiguity simply produces ambiguity at machine speed. A prompt is not a constitution. Permission is not authority. A capable model is not, by capability alone, a legitimate organizational actor. Self-management practitioners begin closer to the real problem because we have already spent years making the invisible structure of work visible. ## What self-management practitioners already know Holacracy, Sociocracy, Teal, and the wider self-management movement are not interchangeable. Holacracy and Sociocracy offer explicit governance grammars; Teal names a developmental pattern of evolutionary purpose, self-management, and wholeness that Frédéric Laloux helped a much broader audience recognize. Their differences matter. Their shared inheritance is a set of practiced instincts that conventional management still struggles to make operational. **The work can be separated from the worker.** Roles describe functional work, purpose, authority, and accountabilities without reducing the person filling them to a job. That makes work portable across humans and Minds while preserving the distinction between role and soul. **Authority can live in structure.** Constitutions, circles, roles, domains, policies, and governance records allow legitimate authority to move beyond personality and proximity. Artificial agents need this structural answer to what they may do, for what purpose, within which bounds, and where human judgment remains required. This is the difference between autonomy and drift. **Tensions, objections, and drivers are information.** These practices teach the system to metabolize discrepancy. My definition makes the connection explicit: > **Intelligence is the rate at which a system metabolizes the tension it encounters into purposeful change.** Capability alone remains neutral; a cancer cell is highly capable at pursuing a local purpose the organism cannot legitimately endorse. So the second definition carries the governance burden: > **Governed intelligence is intelligence in service of a purpose legitimate at its own scale.** **Governance and operations are different kinds of work.** An agent may operate broadly within a role while lacking authority to expand that role, rewrite its constraints, enlarge its budget, or change the policy governing its actions. Operational agency can be fast. Structural change remains explicit, reviewable, and legitimate. **Purpose and boundaries create local freedom.** Purpose acts as a relevance filter; boundaries give intelligence somewhere coherent to act. Rules in service of freedom. The paradox becomes practical once you have lived it. **Clarity is an operating condition.** Roles, policies, domains, decisions, and tensions must be visible enough for distributed actors to coordinate. In an Algorithmic Organization, that hard-won transparency becomes machine legibility. An authority that exists only in someone's head cannot reliably govern a Mind. **Good objections improve the whole.** Mixed human–AI systems need agents able to surface tensions, reveal inconsistencies, and challenge assumptions without confusing computational fluency with moral standing or final authority. Standing is a governance grant, not a metaphysical verdict. Self-management has been teaching us to release distributed agency while retaining coherence. That is no longer a specialist concern. It is the alignment problem inside the organization. ## The inheritance is real. It is not sufficient. There is a tempting but unhelpful conclusion available here: *We already solved this.* We did not. Holacracy is not an Algorithmic Organization. Sociocracy is not an Algorithmic Organization. A self-managing culture does not become agent-ready by opening an API account and assigning a chatbot to a circle. If we down-translate the new frontier into our familiar frameworks, we will lose the frontier while congratulating ourselves on recognizing it. The inherited governance kernel must be extended. An intelligence-bearing organization needs explicit ways to establish an agent’s identity, purpose, capabilities, constraints, and governance standing. It needs memory that can be inspected and corrected; receipts for consequential handoffs; budgets for money, compute, tokens, and attention; escalation paths; versioned policies; proof standards; and hard prohibitions against unauthorized self-expansion. It needs legal bridges between constitutional governance and the entities that hold assets and liability. It needs human judgment at the points where legitimacy cannot be delegated to pattern recognition. Most of all, it needs a more precise human role. I call that role the **Orchestrator**: a human who convenes, stewards, and calibrates a field of intelligence in service of purpose. The old manager pushed work through people. The Orchestrator holds purpose, senses tensions, stewards authority, allocates scarce attention and resources, calibrates Minds, and keeps the system honest. This is not “managing agents” with newer nouns. It is a different relationship to intelligence. And here again, self-management practitioners have an advantage. You have already practiced the underlying moves: - role, not person; - purpose, not preference; - tension, not failure; - objection, not obstruction; - authority, not charisma; - governance, not control; - autonomy, not absence of structure. These are not the whole practice of orchestration. They are unusually good soil for it. ## From a specialist movement to a field of institutional R&D For years, self-management has often been treated as an organizational specialty, a niche within a niche: compelling to a subset of founders, consultants, cooperatives, intentional communities, and teams willing to tolerate unusual meeting vocabulary. I think that frame must now expand. The movement has functioned as a distributed research program into post-managerial coordination. Its living laboratories have been imperfect and sometimes exhausting. Some implementations moved authority; others mostly changed the meeting format. Some produced real sovereignty; others concealed a founder hierarchy beneath more elegant language. Many of us have accumulated scar tissue along with insight. Good. Scar tissue is data—provided the injury was survivable. The next generation of organizations will need people fluent in authority without dominance, structure without centralization, autonomy without drift, and continuous change without perpetual reorganization. It will need practitioners who understand that sensing a tension is not insubordination, that an objection can be an act of service, and that purpose must do real decision work, not simply inspire. That is this community. The opportunity is not to deify or defend Holacracy, Sociocracy, or self-management as the final answer. It is to bring forward what the practices have taught us, allow those teachings to evolve, and help build the governance layer for intelligence-integrated organizations. No victory lap is required. The future is generally unimpressed by those. ## An invitation to lead The Age of Intelligence will not reduce the importance of human sovereignty. It will make sovereignty non-delegable, non-negotiable. As artificial intelligence gains memory, tools, continuity, and operational agency, humans will be asked to do less of the work that once made us feel indispensable. Our contribution will move upstream: holding purpose, making legitimacy judgments, designing authority, attending to consequences, and deciding what kinds of systems deserve to exist. That is not a diminishment of the human role. It is a clarification. Those who have practiced self-management are unusually prepared for it—not because we have escaped hierarchy, ego, ambiguity, or power, but because we have learned to see them as design material for what's next. We have learned that empowerment is systemic, that autonomy needs structure, and that purpose is only real when it can govern a choice. The Algorithmic Organization will require new runtimes, memory systems, legal forms, agent protocols, and technical safeguards. But beneath all of them sits an older question: can intelligence be given enough freedom to act while remaining in legitimate service to the whole? Holacracy and Sociocracy have been working on that question all along. So has everyone who has tried, seriously and sometimes painfully, to move authority out of the person and into a transparent practice. The Age of Intelligence did not make that work quaint. It revealed what it was for: a prerequisite for what comes next. **You’ve been training for this all along. It is time to step forward and model what is now possible.** If you are already practicing Holacracy or Sociocracy, you have already done the hard part: you have learned to map the organization functionally and move authority into explicit practice. Start with one Circle—with or without Sub-Circles—and pilot governed agents as role-fillers. Give them a defined purpose, accountabilities, domains, capability envelope, constraints, and escalation paths. Keep human judgment where legitimacy requires it. Keep receipts. Let tensions update the structure. If you are working in a conventional organization, start with a bounded pilot: one team, department, or consequential workflow. Map it functionally first—purpose, roles, authority, policies, and boundaries—then introduce governed agents inside that structure. Draw on experienced Holacracy or Sociocracy practitioners if useful, or build the map yourself. The point is not to automate the org chart. It is to make authority explicit enough for intelligence to act without drifting. However you enter, we are here to support the journey. [Go free or go fast with PowerShift® Intelligence](https://go.powershift.io): adopt the open constitutional layer yourself, or work with us to form and accelerate an Algorithmic Organization. *Tension is fuel. Purpose sets the bounds.* --- ### Put it into practice - [The PowerShift® Constitution v1.1](https://github.com/powershift-io/powershift-stack/blob/main/constitution/PowerShift_Constitution_v1.1.md) — the fork of the Holacracy Constitution that makes the article’s central moves operational: human and artificial role-fillers, the Orchestrator, capability envelopes, hard constraints, governance standing, and human-intelligence ratification. - [The PowerShift® Stack](https://github.com/powershift-io/powershift-stack) — the full open governance operating system for intelligence-integrated organizations: constitutional, legal, agent, memory, orchestration, and assurance layers. - [The Orchestrator’s Field Guide](https://surface.powershift.io/orchestrator-guide/) — a practical guide to convening, stewarding, and calibrating fields of human and artificial intelligence. ### Follow the lineage - [From Holacracy to Algorithmic Organization](https://surface.powershift.io/lineage/index.html) — the public lineage from self-management and for-purpose enterprise to governed intelligence. - [Holacracy Constitution v5.0](https://www.holacracy.org/constitution/5-0/) and [Holacracy’s distributed-authority overview](https://www.holacracy.org/how-it-works/distributed-authority/) — foundational sources for the constitutional transfer of authority. - [Sociocracy: basic concepts and principles](https://www.sociocracyforall.org/sociocracy/) from Sociocracy For All. - Sociocracy 3.0 patterns for [circles](https://patterns.sociocracy30.org/circle.html), [roles](https://patterns.sociocracy30.org/role.html), and [consent decision-making](https://patterns.sociocracy30.org/consent-decision-making.html).