01 / Provenance
Follow the source.
Connect material values to documents, records, or transactions. Distinguish what was extracted, mapped, and derived so a consuming agent can inspect the source and derivation.
Trust / Healthcare AI governance
Trust is in
the response.
Healthcare AI needs a dependable foundation: shared meaning, traceable evidence, explicit time, and clear authority. These are the principles behind ANY’s contract layer.
Governance principles
Context that stays
with the answer.
Our vision makes evidence, time, uncertainty, and authority part of every healthcare contract—so governance can travel with the facts as agents, models, and workflows change.
02 / Time
Keep both clocks.
When a fact applied and when its evidence was observed are different questions. Carry both times with the result so each workflow can interpret the fact in context.
03 / Uncertainty
Leave room for unknown.
Missing evidence does not establish a negative answer. Conflicting sources should preserve the disagreement and identify what needs review.
04 / Authority
Declare the boundary.
Make permitted uses, decision rights, and review points explicit. Shared facts should support different agents while preserving accountability in each workflow.
Meaning before automation
Unknown
is a valid result.
A determination describes what the evidence establishes. An execution status describes what happened when the contract ran. Keep those meanings separate.
knownEvidence supports the determination. That can include an explicitly supported negative answer.
unknownAvailable evidence cannot establish the answer. It must not silently become false.
conflictingSources disagree. The disagreement remains visible for review.
unresolvedExecution did not produce a supported domain result. Do not interpret it as a denial.
failedAn execution failure is a technical or access outcome, not a healthcare determination.
Illustrative vocabulary for contract design. Published artifact maturity and deployment behavior are described below.
Deployment boundaries
Agree the controls
where the work happens.
Production behavior depends on the binding, the customer environment, and the governing agreement.
Access and environment. Establish authorized sources, permitted consumers, tenancy, data location, and the scope of each operation.
Action authority. Evidence supports conclusions; permission to act comes from the deploying organization’s controls and agreements. A documented payment alone establishes neither payment correctness nor authorization to submit or settle a dispute.
Data and operations. Agree retention, subprocessors, audit behavior, operational expectations, and how exceptions reach human review.
Contractual commitments. Verify the controls and agreements required for the specific deployment. This public site is not a certification report or evidence of controls in every environment.
The website is informational and its examples are synthetic. Keep protected health information, medical records, and urgent care requests out of public email.
Inspectable status
Availability is a fact,
not an implication.
Lifecycle
What stage is it in?
Encounter Lifecycle is active. Coverage State is proposed.
Availability
Who can invoke it?
Encounter Lifecycle is private. Coverage State has no executable binding.
Artifact maturity
What can depend on it?
A binding request, a representative response, and a draft schema carry different commitments.
Implementation note: these governance principles describe the platform vision. Coverage State’s determination states are draft. Encounter Lifecycle’s public response and explorer semantics are representative; its current binding returns document extractions for reconciliation in the customer’s workflow. Confirm controls and behavior for each deployment.
Start with a specific workflow
Ask how the contract behaves.
Let’s discuss its sources, uncertainty, review points, and deployment requirements.