The Hybrid Workforce Standard

An AI Employee must hold a defined role, not merely a persona or a system prompt.

The post exists before its occupant. A name, a tone of voice and a set of instructions describe a persona; a role states what result must be produced, with what authority, measured how.

There must be exactly one accountable human manager.

An AI Employee may receive work from many people, but unity of command still applies. A resource with two owners has none, and a resource with no owner is administratively orphaned no matter how well it is integrated technically.

Authority must be explicit, limited and revocable.

What the resource may decide without approval has to be written down before it operates, not inferred afterwards from what it happened to do.

Access must follow least privilege.

Context is capability, but it is also risk surface. More access is not more competence; it is a larger blast radius.

High-risk actions must support human approval.

Proportionality is the rule: the higher the cost of the error and the harder it is to reverse, the more approval the action carries. Separation of duties applies — whoever initiates need not approve.

Every material action must be auditable.

Actor, input, tool, action, approval, result and timestamp. If the sequence cannot be reconstructed after the fact, the organisation cannot learn from a failure or demonstrate compliance.

Performance must be measured by outcomes, quality, risk and cost — never by activity, hours, tokens or message volume.

Speed makes activity look like value. A badly designed process that produced ten errors will produce a hundred once automated, and the dashboard will call it throughput.

Memory must have provenance, scope, retention and deletion rules.

Remembering improves performance and creates exposure at the same time. An undated policy produces answers that are consistent and wrong; a historical exception should not silently become a rule.

The system must know when to escalate rather than improvise.

A resource that produces a convincing answer instead of raising a doubt is more dangerous than a less capable one that is better governed. Every exception needs a destination.

An AI identity must never be used to deceive customers or colleagues about its nature as software.

Giving a system a role, a name and a manager improves coordination. Letting a person believe they are talking to a human being is a different act, and this standard does not permit it.

Humans and AI Employees may share an operational graph while retaining distinct status and rights.

A shared org chart is an administrative convenience, not an ontological claim. Dignity, health, rest, labour rights and belonging are properties of people. Operational anthropomorphism without ontological confusion.

Offboarding must revoke access and transfer or destroy context safely.

Revoke credentials, disable tools, stop schedules and queues, rotate secrets, transfer outstanding work, preserve evidence. Offboarding is as important as onboarding and is almost always skipped.

Every Human ↔ Artificial transition must be evaluable against a baseline and carry explicit rollback criteria.

Without a baseline recorded before the change, the organisation can celebrate an improvement it never made. Rollback criteria written after the results are known are not criteria; they are justification.

An AI Employee must have an identifiable version of model, policies, tools and knowledge base.

Without versioning it is impossible to say what authority existed at a given moment, or which change produced an improvement or a regression.

Not every agent qualifies as an AI Employee. It must clear a threshold of maturity and governance.

This clause is what makes the category worth anything. If the label applies to everything, it distinguishes nothing — and an organisation with twenty excellent agents and zero AI Employees has clarity, not failure.