GLOSSARY
The Six Primitives
Every operation — all of them, always — decomposes into these six irreducible building blocks. Everything else is a composition. Nothing decomposes further. Nothing is missing.
Policy
The rule. Why something must happen. Policies are declarative — they state what must be true, not how to make it true. Policy is constraint, not intelligence.
Policies answer the questions your system must know from day one:
- "Can this operator take this job?" → assignment policy
- "Does this step require a photo?" → spot check policy
- "How many kids fit in a party slot?" → capacity policy
- "Does this critical finding need escalation?" → escalation policy
Procedure
The steps. How to do the thing. A procedure is a sequence of actions that satisfies a policy. Procedures are specific — they name the thing, the steps, and the order.
"Property Inspection — Unit Turnover" — condition check, photos, repairs flagged, manager sign-off, keys.
"Lawn Service — Full Visit" — arrive, assess, mow, edge, blow, photograph, complete.
Asset
The thing being acted on. The test for Asset-hood is simple: does work happen on this thing, and does it have state we want to track? If yes, it's an Asset. Not all Assets are physical. Some are ephemeral (a party booking, a job, a legal case). Some are relational (a customer relationship, a vendor contract). Some are purely informational (a brand, a dataset, a configuration). They may exist for a window, then become historical.
A water heater. A property. An emergency room patient. A SaaS subscription. A consulting engagement. A brand guideline. A production dataset.
Actor
Who — or what — does the work. An Actor is any entity that can be delegated to, holds qualifications, has availability, and is attributed in the ledger. The biological test is not the test. The operational test is.
Humans are Actors. So are non-human Actors: software agents, lab systems, kitchen lines, channel sessions, automated dispensers.
If an entity holds capabilities, can accept or refuse work, and writes to the ledger as applied_by, it's an Actor.
If not, it's infrastructure, a Trigger, or a side-effect — not an Actor.
Three role archetypes recur: the Actor doing the work, the Supervisor accountable for it, and the Customer it's done for.
| Context | Doer | Supervisor | Customer |
|---|---|---|---|
| Property | Steward | Manager | Homeowner |
| ER | Clinician | Charge Nurse | Patient |
| Pharmacy | Dispenser (system) | Pharmacist | Patient |
| Lawn Care | Provider | Platform Ops | Homeowner |
Trigger
What causes work to happen. Triggers don't just start individual procedures — they start chains of work. Trigger + Policy together are the operational brain.
Scheduled — Monday 8am. Quarterly. 30 minutes before close.
Condition-based — Member count > 45. Provider rating < 3.5. Coverage gap detected.
Occurrence-based — Customer books service. Provider marks complete. Injury reported.
Ledger
The record. Proof it happened. Every execution of every procedure writes to the ledger. The ledger is append-only. The ledger is the moat.
The ledger also feeds back into operations: policies can query the ledger to make decisions ("3 complaints in 30 days" requires reading historical entries). Over time, the ledger becomes the training set for cross-business intelligence.
Inferred
The opposite of authored. An inferred operation runs on policies that exist only because a senior person "just knows," procedures that vary by who's on shift, and records that depend on someone remembering to write things down.
Inferred operations work — until the person who holds the knowledge leaves, scales, or makes a mistake. Inferred systems also break under AI: a model trained on inferred behavior learns the habits, not the rules.
Qualifications
The capabilities or authorizations an Actor holds that gate what work can be assigned to it. A human's qualifications are certifications, training, clearances, licenses. A system's qualifications are the drugs a dispenser can release, the assays a lab machine can run, the models a service can invoke, the queries a channel session can answer.
Qualifications answer: "is this Actor allowed to do this work?" If the answer isn't explicit, assignment is guessing.
Availability
The state that lets an Actor accept or refuse work right now. For humans: schedule, capacity, on-call status, time off. For systems: uptime, maintenance windows, rate limits, queue depth.
Availability is the live counterpart to Qualifications. Qualifications are can. Availability is now. Assignment without availability is wishful thinking.
Kill Switch Test
A diagnostic from Principle 9. "What happens if you turn off the AI?" If your operation stops, the AI isn't assisting your operation — it is your operation, and you've built a dependency, not a system.
An authored system passes the Kill Switch Test: Policies, Procedures, Assets, Actors, Triggers, and the Ledger remain intact and operable when the perception layer is off. AI adds speed. It is not the foundation.
The Flow
Triggers signal that something happened (or should happen).
Policies decide what to do about it.
Procedures define the work and execute the steps.
Actors do the work, on Assets.
The Ledger records everything.
The Ledger feeds back into Policy decisions.This is not a linear pipeline. It's a loop. Triggers fire. Policies evaluate constraints. The engine runs procedures. The ledger records outcomes. Those outcomes feed back into future policy decisions.