In 1937, economist Ronald Coase asked in “The Nature of the Firm”: if markets are so efficient, why do firms exist at all? His answer was that using the market isn’t free; discovering prices, negotiating terms, and arranging a contract for each exchange — these all cost something. Work moves inside a firm wherever directing it there costs less, the firm growing until those costs meet.
But what happens to the firm as AI makes coordination radically cheaper? The Exponential Organizations community has spent more than a decade exploring how technology reshapes organizations. Salim Ismail, Michael Malone, and Yuri van Geest’s Exponential Organizations (2014) described companies whose reach grows far beyond their headcount. OpenExO grew around that work, with coaches helping organizations put its ideas into practice.
Now The Organizational Singularity, led by Ismail, extends the argument into ExO 3.0: an organization built around an Intelligence Stack, with governance running through it. More execution passes to agents; people retain responsibility for purpose, judgment, and consequential decisions. The firm becomes, increasingly, an accountability shell.
That picture has an immediate implementation requirement. Authority must be expressed in a form that people can understand and machines can enforce.
At Engaging, we’ve built around precisely that requirement. The result is a close fit between several ExO ideas and the everyday workings of Engaging OS.
The original ExO framework grouped ten attributes into SCALE and IDEAS. ExO 3.0 introduces DRIVE and SHAPE, but the earlier attributes remain a useful way to see the fit:
ExO attribute | Its expression in Engaging OS |
|---|---|
Autonomy | Positions carry explicit authority to act. They get instantiated by Roles at various bodies; people or agents in turn hold roles. |
Dashboards | The resulting authority determines which items appear in each Role’s Console. |
Algorithms | Agents hold Roles and act through the same permission system as people. |
Staff on Demand | Employees, contractors, and agents can fill Positions. |
Interfaces | Outside contributors enter through defined Positions and work on the same items. |
Experimentation | A process can be modeled and running in a single session. |
These connections reinforce one another. The permission that gives someone autonomy also populates their Dashboard. The Position that accommodates a contractor also accommodates an agent. The Operations that run the work also define its limits.
Consider them in turn.
Autonomy and Dashboards
ExO’s Autonomy attribute describes self-organizing teams with authority pushed down to them. Dashboards give those teams shared, current information to steer by. Together, they let work proceed without waiting for a manager to direct each step.
In Engaging OS, authority and visibility come from the same definition.
A Position holds Operations: the right to take a particular action on an item in a particular status. A Role assigns a person or agent to that Position at a Body for a defined period. Its holder’s Console shows the items that those Operations let persons and agents act on now.
A draft moves to submitted. It leaves the drafter’s actionable list and appears on the approver’s. Once approved, it becomes available to whoever sends it.
Nobody routed it; its status did.
Biologists call coordination through changes in a shared environment stigmergy. An ant leaves a pheromone trail; a Wikipedia editor leaves an edit for others to inspect and develop. Here, the trace is an item’s status. Each change makes the next authorized action available.
The process needs to be designed, but its routine handoffs need no further arrangement. Autonomy supplies the authority; the Dashboard shows where to exercise it.
This also makes the Statuses and Operations the workflow. Publish on draft and redraft on published form a cycle. Add submit and approve, and you have an approval chain. There is no need to produce a separate routing diagram to keep aligned with the permissions, because the specs, complete enough to be executable, are also complete enough to generate on-the-fly business process (BPMN) diagrams.
Note that this means the organization is not defined by who has authority over whom; rather, it is about who can do what to various objects, among which are Persons. Autonomy as a value is baked in to the Engaging OS model of what an organization actually is.
Algorithms and Staff on Demand
ExO’s Algorithms and Staff on Demand expand what an organization can accomplish with its permanent staff. Software and outside talent take on work the firm would once have hired people to perform internally.
Engaging OS gives each contributor a defined place in the operation. A Position can be filled by an employee, a contractor, or an agent. The holder may change; the Position’s authority remains explicit.
For an agent, that authority is its permission envelope.
Take sales outreach. Three Positions participate: the Drafter, the Approver, and the Sender. The Drafter and Sender are agents; the Approver is a person.
The Drafter’s envelope is small:
It may draft outreach for a qualified Lead.
It may submit the draft for approval.
It may revise and resubmit a returned draft.
The Drafter finds a qualified Lead in its Console, writes the outreach, and submits it. The item leaves its actionable list and enters the Approver’s queue.
Ask the Drafter to approve its own work and it can’t. That Operation isn’t available to its Position. The refusal comes from the platform, rather than an instruction the agent must remember to obey.
The Approver returns the draft with a requested change. The Drafter revises and resubmits it. The Approver approves. Only then does the outreach become available to the Sender.
Each step records who acted, what changed, and when. The organization gets agent execution, a human review queue, separation of authority, and a record of the work through one set of operational definitions.
Returning a draft allows revision. Sending it has an external consequence. The distinction belongs in the permissions: a person holds approval authority, while the Sender executes the approved action.
The accountability shell acquires a working interior.
Interfaces
ExO’s Interfaces connect outside resources to the organization’s internal work. Without them, access to a crowd or a pool of specialists can create another coordination job.
Among the core values of Engaging OS is Frank Lloyd Wright’s notion of organic architecture, where the outside and inside are clearly defined as different, yet each is of the other.
In Engaging OS, an outside contributor enters through a Position. Their Operations define what they can see and do, and their work lands on the same items used by everyone else.
Interfacing with the system occurs one of two ways: humans get screens, while AIs get the the industry-standard MCP protocol.
A contractor’s submission can become an employee’s approval item. An agent can prepare something for an external specialist to review. The shared process continues across the boundary because the permissions describe each participant’s part in it.
The organization’s edge becomes operationally precise: these contributors may take these actions on these items under these conditions.
Experimentation
ExO treats Experimentation as an organizational habit: try an idea, observe what happens, and adjust.
Experimentation is easier when changing a process begins with sentences the business can read.
Engaging OS uses BON/D—Business Operations Notation/Development—to express Operations. Its questions make the actor, action, item, conditions, and context explicit:
A Guest can book a Room, provided that Room has availability on that date.
Anyone involved can read that, challenge it, and agree on the intended rule. The sentence specifies the Operation the platform enforces.
Bring a bounded process to a modeling session and it can be running before the session ends. People can then inspect actual work: what appears in each Console, where an item waits, whether a Position has enough authority, and whether an approval is necessary.
The model becomes something to try.
We work this way ourselves. Our Domain Analyst agent interviews people, records the terms they use, and drafts initiatives. A System Architect, human or AI, turns that understanding into BON/D sentences and configures the system. Discovery, modeling, and operation stay close enough to inform one another.
Purpose and the people around it
ExO begins with a Massive Transformative Purpose. Purpose attracts people, gives their effort direction, and helps a community form around the organization. While the domain analysis discussions can be inspiring, software obviously cannot supply that commitment.
Nor does assigning someone a Position create trust, engagement, or a productive community. Engaging OS can give contributors a way to act once those relationships exist. Leaders and coaches still have to build them.
ExO supplies a language for the organization people want to become. Its coaches help them make the choices that entails. Engaging OS makes several of those choices executable: authority becomes Operations, responsibility becomes a Role, and a change in an item’s status brings the next participant into the work.
The match is particularly useful as agents join the organization. Human and agent contributions can share a process while retaining distinct permissions. Autonomy can expand without leaving its boundaries implicit.
Coase explained the firm through the cost of coordination. As that cost falls, the firm still needs a clear account of who may act and who answers for the result.
ExO describes the emerging organization; Engaging OS provides the backbone for its realization.
For more about the rise of the Exponential Organization, see:
“The Rise of the Exponential Organization with Peter Diamandis and Salim Ismail” at the A16Z podcast