DUCIMUSThe one-human company changes the shape of small business.
One human can increasingly coordinate work that used to require several people. Specialized agents can research, monitor, communicate, publish, buy, organize, analyze, and handle routine operations. An Executive Agent can sit above them, carrying context across the company and coordinating work on behalf of the owner.
That creates leverage.
It also creates structure.
And once agents become part of the structure of the business, productivity is no longer the only thing worth measuring.
The business has to ask whether that structure is actually under its control.
That is where the Business Continuity Test for the Agentic Company begins.
The test has four parts:
Identity. Authority. Portability. Succession.
Together, they answer a larger question:
Does the business control its agentic infrastructure, or does the infrastructure quietly control the continuity of the business?
1. Identity: Does the business know who it is without the provider?
Every company needs an identity that survives its tools.
That sounds obvious because we already expect it from ordinary software. A business does not become a new company because it changes accounting platforms, email providers, or website hosts.
Agentic systems complicate that because an Executive Agent may eventually become deeply woven into the company's operating identity.
Other systems may recognize it.
Vendors may interact with it.
Customers may encounter it.
Specialized agents may receive authority through it.
Years of operating history may become associated with its identifier.
If that identity exists only inside one provider, then the provider may be holding more than a software account. It may be holding the identity through which part of the business is recognized operationally.
That is why the first test is simple:
If the provider disappeared tomorrow, could the business still prove who it is?
The business should ideally sit above the agent.
The agent should sit above the model or runtime.
The hierarchy should look conceptually like this:
Business identity → Executive Agent → specialist agents → models, tools, and runtimes
The lower layers can change without necessarily destroying the identity above them, provided the business has preserved enough identity, authority, records, and standing outside those layers to make replacement possible.
When that separation exists, an Executive Agent can be replaced without requiring the company to reinvent itself.
When it does not, changing infrastructure may begin to look more like rebuilding the organization.
2. Authority: Can the business explain who gave every agent permission to act?
Agents become much more consequential when they stop merely suggesting actions and begin performing them.
At that point, access is not enough.
A business needs to understand authority.
An agent may technically have access to email while lacking authority to accept a contract. It may be able to reach a payment system while being limited to preparing transactions rather than releasing them. An Executive Agent may coordinate a launch without being authorized to create an unlimited advertising budget.
Those distinctions have to survive delegation.
If the human authorizes the Executive Agent, and the Executive Agent assigns work to a specialist agent, the specialist should not somehow gain more authority than the Executive Agent received.
The business should be able to trace the chain.
Who authorized the action?
What exactly was authorized?
What limits were attached?
Could that authority be delegated again?
When did it expire?
Was it still valid when the action occurred?
How can it be revoked?
How are consequences already created remediated?
Revocation controls future action. Remediation addresses consequences already created.
This is where the agentic company begins to look less like a stack of apps and more like an organization.
It has roles.
It has boundaries.
It has delegation.
It has accountability.
And because one human may sit at the top of all of it, those boundaries need to be clearer rather than looser.
A business that cannot explain where agent authority came from does not fully understand its own operating structure.
3. Portability: Can the company move without losing the parts that matter?
Data portability is familiar.
Agent portability is harder.
An Executive Agent may eventually contain or coordinate much more than information. It may have an established identity, history, credentials, permissions, relationships, reputation, workflow state, and recognized standing with other systems.
Moving the files does not guarantee that those things move.
That is why portability needs to be tested in layers.
Can the instructions move?
Can the memory move?
Can the identity move?
Can credentials be reissued?
Can the agent prove that it still represents the same business?
Can outside systems recognize its successor?
Can prior actions remain auditable?
Can the business preserve relationships that were built through the old identity?
The most important distinction is between data portability and standing portability.
I’m using standing portability descriptively as well; it is not the name of an established technical standard.
The business may be able to export everything it technically owns while still losing the position it earned inside a network.
Creators already understand this intuitively.
You can download your content without downloading the algorithmic relationship that formed around it.
Businesses may eventually discover the same thing with agents.
They may be able to export the agent's memory without exporting the trust that other agents, vendors, and platforms built around the old identity.
That is why the portability question cannot stop at:
Can I download my data?
It eventually becomes:
Can I export my standing?
The answer may never be completely yes.
The business still needs to know which pieces can travel and which are attached to the infrastructure it is leaving.
4. Succession: What happens when the current controller is gone?
Every system looks stable while all of its important parts are present.
Continuity becomes visible when one disappears.
The one-human company has two especially important succession problems.
The first is human succession.
If the owner becomes incapacitated or dies, the company may still have functioning agents, active credentials, scheduled transactions, customer conversations, vendor commitments, and automated workflows.
Technical continuity may remain intact while governance becomes uncertain.
The Executive Agent may know more about the business than anyone else.
That does not mean it gets to decide that it is now in charge.
Authority has to move through a legitimate succession path.
That may involve an executor, trustee, surviving member, manager, attorney-in-fact, or another person whose authority comes from the business and legal structure rather than from possession of a password.
The second problem is agent succession.
I’m using agent succession here as a descriptive term, not the name of an established legal doctrine.
If the Executive Agent disappears because the provider fails, the account is lost, the product is discontinued, or the business chooses to leave, another agent may need to take over the same role.
That replacement should not have to become a completely new business identity.
The company should be able to say:
The old Executive Agent is no longer authorized. This is the legitimate successor. The company is the same.
That is a continuity problem, not merely a migration problem.
The four parts are connected.
Identity without authority tells the system who an agent is but not what it may do.
Authority without portability may leave the business unable to move.
Portability without succession may preserve technical assets while leaving nobody legitimately authorized to control them.
Succession without independent identity may transfer control over something that still cannot survive outside its original provider.
The four dimensions therefore reinforce one another.
A resilient agentic business should be able to answer:
Who are we?
Who is allowed to act?
What can move?
Who takes over when someone or something cannot continue?
Those are business questions before they are technical questions.
The simplest continuity test uses two scenarios.
The first scenario removes the human.
Ask:
If I disappear, who takes control?
Not who knows the passwords.
Not who understands the software.
Who has legitimate authority to govern the business, review active operations, control the Executive Agent, rotate credentials, revoke permissions, and appoint whatever comes next?
The second scenario removes the provider.
Ask:
If my AI provider disappears, does the business still know who it is?
Can the business prove its own identity?
Can another Executive Agent take over?
Can the company preserve enough operating history to continue?
Can external systems recognize the transition?
Can the business recover without rebuilding its organizational structure from scratch?
Those two scenarios expose the architecture very quickly.
A business can pass this test in degrees.
This is not a pass-or-fail certification.
A small business may reasonably depend on providers.
Almost every modern company does.
The goal is not to eliminate dependency.
The goal is to understand which dependencies are merely inconvenient and which dependencies threaten organizational continuity.
If losing a provider means changing a workflow, that is ordinary switching cost.
If losing a provider means losing the identity through which the business is recognized, the authority structure through which agents act, years of accumulated operating history, and the relationships the company has built with other systems, that is something deeper.
That is where rented organizational structure becomes visible.
The business may still own itself legally.
It may still own its brand.
It may still own its customer relationships on paper.
But part of the machinery through which the organization actually functions may exist inside someone else's identity, permission, and execution environment.
That is not automatically bad.
It should be intentional.
The question is not whether agents belong inside the company.
They increasingly will.
The question is how they belong there.
An Executive Agent can become valuable precisely because it carries context, coordinates work, maintains continuity, and allows one human to govern more productive capacity.
Removing all of that would defeat the point.
The better design is to make the agent important without making it irreplaceable.
The Executive Agent should be able to know the business without being the only place the business knows itself.
It should be able to build relationships without becoming the sole owner of those relationships.
It should be able to exercise authority without becoming the source of that authority.
It should be able to coordinate the organization without becoming the organization.
That is the line.
The one-human company is not really about replacing employees.
That framing is too narrow.
The deeper shift is that organizational capacity is beginning to detach from human headcount.
One human may eventually govern a business containing many operational entities, each with roles, permissions, histories, and responsibilities.
That creates a new kind of small company.
The human may do less execution while carrying more responsibility for orientation, judgment, boundaries, and governance.
That is why the important measure is not how many agents the owner can deploy.
It is how much agentic capacity the owner can understand and govern without losing control of the business underneath it.
The real advantage of the one-human company is not maximum automation.
It is controlled leverage.
And controlled leverage requires the business to remain governable through the loss or replacement of any one agent, model, or provider, and to have a defined succession path when the human authority can no longer act.
That is the continuity test.
If the human disappears, the business should know who can take control.
If the Executive Agent disappears, the business should know how to replace it.
If the provider disappears, the business should still know who it is.
When those things are true, agents are working inside the company.
When they are not, part of the company may be living inside the agents.