Corporations Are People, My Friend

In August 2011, Mitt Romney said, “Corporations are people, my friend.” The line became a punchline, but the underlying legal point was ordinary and correct.

Corporations are juridical persons. They can sign contracts, own property, sue and be sued, hold bank accounts, and bear liability. That legal fiction is what lets a company persist as an entity rather than dissolve into the individual people inside it.

The obvious limit is that the corporation has never had agency of its own. People inside the entity think, act, and decide on its behalf.

Corporate agency before software

A corporation is a principal that acts through human agents: officers, directors, employees, and contractors. The board passes a resolution and a person carries it out. The bylaws define a process and a person executes it. The entity signs a contract because a person signs for it.

Corporate governance has always been built around that fact. Oversight, delegation, signatures, approvals, and audit controls exist because the entity itself is inert and must rely on the people acting for it.

When governance fails, the failure usually sits somewhere in that delegation chain. The entity did not decide to commit fraud. The people acting for it did.

Enter the agentic corporation

A corporation can now operate through software agents that read its records, apply its bylaws, execute authorized actions, manage treasury workflows, and prepare filings.

That is the agentic corporation: a legal entity with AI agents operating under its authority and within the limits set by its governing documents.

The underlying pieces already exist. AI agents can read and write structured data, follow rule sets, call APIs, process documents, and execute multi-step workflows. What matters is the governance layer that binds those capabilities to actual corporate authority.

Consider an agent with scopes like formation_read, treasury_write, and equity_read. Those are not just API permissions. They are a delegation of authority expressed in code. The agent can read formation documents because the corporation authorized it to. It can write treasury entries because the authority model permits it.

When an agent inherits scopes from a parent agent, the chain of delegation becomes explicit, auditable, and mechanically enforced.

Better principals, better agents

The usual fear is that AI agents will act outside their authority or make expensive mistakes. Those concerns are real. They are also familiar. Corporate law and internal controls already assume that agents, human or software, need boundaries.

The important difference is that software agents can be constrained in ways humans cannot. A human officer with treasury access can still misuse it. A software agent with treasury_write scope can be prevented from touching equity records at the infrastructure level.

Scoped, time-limited, auditable delegation is the point of the model. Every action traces back to a grant of authority. Tokens expire. Scope boundaries are enforced.

Taken seriously, corporate personhood becomes a design constraint. The corporation has its own identity, its own records, its own authority, and now a practical mechanism for acting through software.

The version-controlled corporation

One more piece makes this operational rather than theoretical. In an agentic corporation, formation documents, equity records, resolutions, and treasury state live in version-controlled storage. Every change is a commit with provenance.

When an agent acts on behalf of the corporation, the system records who authorized it, what scopes were granted, what changed, and when it happened. That history lives in a cryptographically linked chain of commits rather than in an editable application log.

The result is a corporate record that is easier to verify, easier to audit, and harder to separate from the actions it describes.

People all the way down

Romney was right in the legal sense. Corporations are legal persons. For centuries, their ability to act has been borrowed entirely from humans.

The agentic corporation does not remove people from governance. Directors still set strategy. Shareholders still vote. Officers still make judgment calls. What changes is the operating layer. Maintaining records, enforcing bylaws, executing authorized resolutions, and managing compliance no longer require a person to touch every step.

For the first time, the legal person can operate through a software system designed to follow its own rules.