Term reference
Agentic Ontology of Work 2.0.0
All terms in the aow: namespace, https://w3id.org/aow#. Each term identifier resolves to its entry on this page.
| Ontology IRI | https://w3id.org/aow |
| This version | https://w3id.org/aow/2.0.0 |
| Formats | Turtle · JSON-LD · JSON-LD context · SHACL shapes · JSON Schema |
| Released | 2026-09-24 |
| Creator | Manish Garg, Skan.ai |
| License | Apache 2.0 |
Classes
aow:WorkItem Work Item Perception
The unit of business work that the rest of the ontology is about: a claim, an application, an order, a case, a ticket.
aow:Signal Signal Perception
A raw, time-stamped trace emitted by a system, a person, or the environment: an event, a log line, a user-interface interaction, a document arriving, a state change, a metric sample.
aow:Observation Observation Perception
An interpreted fact about a Work Item, Actor, or Action, derived from one or more Signals and carrying a Confidence.
aow:Objective Objective Cognition
A business-level goal expressed in strategic or operational terms.
aow:Intent Intent Cognition
A structured, actionable goal derived from an Objective and interpretable by agents, orchestrators, and people.
aow:Context Context Cognition
The structured situational information that gives relevance to an Intent.
aow:Policy Policy Cognition
Declarative constraints defining permissible, required, or restricted behavior.
aow:Plan Plan Cognition
An ordered or partially ordered set of Tasks intended to fulfill an Intent, produced by an Orchestrator or other Actor.
aow:Task Task Execution
A unit of assigned work within a Plan, performed by an Actor filling a Role.
aow:Actor Actor Execution
Anything that can be assigned Tasks, perform Actions, and be held to account for them: an Agent, a person, an Orchestrator, or a Guardian.
aow:Agent Agent Execution
A software Actor that interprets Intents and Tasks and acts on them using Skills, within the autonomy it has been cleared for. A kind of Actor.
aow:HumanActor Human Actor Execution
A person who performs, approves, reviews, or oversees work. A kind of Actor.
aow:Orchestrator Orchestrator Execution
A coordinating Actor that turns Intents into Plans, assigns Tasks to Actors, and manages sequencing and exceptions. A kind of Actor.
aow:Guardian Guardian Assurance
The oversight Actor that enforces Policy and Assurance Levels, monitors behavior, and escalates to people when needed. A kind of Actor.
aow:Role Role Execution
A named set of responsibilities and authority that an Actor can fill, such as "claims adjuster" or "eligibility reviewer".
aow:Action Action Execution
A single execution event in which an Actor invokes a Skill, in service of a Task, at a point in time.
aow:Confidence Confidence Assurance
A model-, rule-, or human-derived measure of certainty about an Observation, a Result, or an Action.
aow:AssuranceLevel Assurance Level Assurance
A governance-set level of permitted autonomy, with the human oversight and evidence it requires.
aow:Outcome Outcome Assurance
The aggregated business impact of one or more Results, measured against an Objective.
aow:Feedback Feedback Assurance
Structured information, from people or systems, used to improve future behavior.
aow:Memory Memory Assurance
Persistent knowledge accumulated from experience, used to inform future Context.
aow:OperationalMemory Operational Memory Assurance
Short-lived, execution-oriented memory scoped to a session, Task, or Work Item. A kind of Memory.
aow:KnowledgeBase Knowledge Base Assurance
Long-lived memory of generalized patterns, rules, and precedents. A kind of Memory.
Relationships
aow:servesObjective serves objective
The Objective an Intent was derived from.
From Intent to Objective. JSON key serves_objective.
aow:shapedBy shaped by
Context that gives the Intent or Task its situational meaning.
From Intent, Task to Context (many). JSON key shaped_by.
aow:drawsOn draws on
Memory and Observations the Context was assembled from.
From Context to Memory, Observation (many). JSON key draws_on.
aow:concerns concerns
The Work Item the Intent, Task, or Action is about.
From Intent, Task, Action to Work Item (many). JSON key concerns.
aow:constrainedBy constrained by
Policies that apply.
From Intent, Plan, Task, Actor, Skill, Role to Policy (many). JSON key constrained_by.
aow:requiresAssurance requires assurance
The highest Assurance Level at which this work may be done. The effective level is the lowest of the Intent's, the Task's, the Skill's, and the performing Agent's autonomy_level.
From Intent, Task, Skill to Assurance Level. JSON key requires_assurance.
aow:realizes realizes
The Intent the Plan is meant to fulfill.
From Plan to Intent. JSON key realizes.
aow:plannedBy planned by
The Actor (usually an Orchestrator) that produced the Plan.
From Plan to Actor. JSON key planned_by.
aow:hasTask has task
The Tasks that make up the Plan.
From Plan to Task (many). JSON key has_task.
aow:dependsOn depends on
Tasks that must complete before this one can start.
From Task to Task (many). JSON key depends_on.
aow:assignedTo assigned to
The Actor responsible for performing the Task.
From Task to Actor. JSON key assigned_to.
aow:requiresRole requires role
The Role an Actor must fill to take the Task.
From Task to Role. JSON key requires_role.
aow:fillsRole fills role
Roles the Actor is authorized to fill.
From Actor to Role (many). JSON key fills_role.
aow:hasSkill has skill
Skills the Actor can invoke.
From Actor to Skill (many). JSON key has_skill.
aow:performedBy performed by
The Actor that carried out the Action.
From Action to Actor. JSON key performed_by. Specializes prov:wasAssociatedWith.
aow:fulfills fulfills
The Task the Action was performed for.
From Action to Task. JSON key fulfills.
aow:invokes invokes
The Skill the Action used.
From Action to Skill. JSON key invokes.
aow:approvedBy approved by
People who confirmed the Action, Plan, or Result before it took effect. Required whenever the effective Assurance Level or a Policy calls for human confirmation.
From Action, Plan, Result to Human Actor (many). JSON key approved_by.
aow:generatedBy generated by
The Action that produced this Result, Signal, or Observation.
From Result, Signal, Observation to Action. JSON key generated_by. Specializes prov:wasGeneratedBy.
aow:derivedFrom derived from
The Signals an Observation was interpreted from.
From Observation to Signal (many). JSON key derived_from. Specializes prov:wasDerivedFrom.
aow:about about
What the Observation describes.
From Observation to Work Item, Actor, Action, Task. JSON key about.
aow:affects affects
Work Items whose state the Result changed.
From Result to Work Item (many). JSON key affects.
aow:hasConfidence has confidence
The Confidence attached to this Result, Observation, Action, or Plan.
From Result, Observation, Action, Plan to Confidence. JSON key has_confidence.
aow:comparedAgainst compared against
The Assurance Level whose confidence threshold the Confidence was checked against.
From Confidence to Assurance Level. JSON key compared_against.
aow:contributesTo contributes to
Outcomes this Result counts toward.
From Result to Outcome (many). JSON key contributes_to.
aow:evaluates evaluates
The Objective the Outcome is measured against.
From Outcome to Objective. JSON key evaluates.
aow:feedbackOn feedback on
What the Feedback is about.
From Feedback to Outcome, Result, Action, Plan (many). JSON key feedback_on.
aow:providedBy provided by
Who gave the Feedback.
From Feedback to Actor. JSON key provided_by.
aow:updates updates
What the Feedback changed.
From Feedback to Policy, Memory, Plan, Skill, Actor, Context (many). JSON key updates.
aow:writesTo writes to
Memory the Actor records its experience in.
From Actor to Memory (many). JSON key writes_to.
aow:enforces enforces
The Policies and Assurance Levels the Guardian enforces.
From Guardian to Policy, Assurance Level (many). JSON key enforces.
aow:oversees oversees
What the Guardian watches.
From Guardian to Actor, Plan, Skill (many). JSON key oversees.
aow:escalatesTo escalates to
Where escalations go.
From Guardian, Policy to Role, Human Actor (many). JSON key escalates_to.
Attributes
Common attributes map to existing properties: label to rdfs:label, description to dcterms:description, version to schema:version, created_at to prov:generatedAtTime, created_by to prov:wasAttributedTo, revision_of to prov:wasRevisionOf.
aow:acceptableRiskBand
The level of residual risk the requester accepts.
On Intent. Type enum. JSON key acceptable_risk_band.
aow:agentRole
What Agents may do at this level.
On Assurance Level. Type text. JSON key agent_role.
aow:artifacts
Documents, records, or state changes the Result produced.
On Result. Type uri-list. JSON key artifacts.
aow:autonomyLevel
The highest Assurance Level the Agent has been cleared to operate at. The Agent may operate at a lower level on a given Task, never a higher one.
On Agent. Type ref. JSON key autonomy_level.
aow:binding
How to invoke it, for example an MCP tool name, an API endpoint, or a UI automation.
On Skill. Type string. JSON key binding.
aow:calibrated
Whether the value has been calibrated against observed accuracy.
On Confidence. Type boolean. JSON key calibrated.
aow:closedAt
When the item left the process, if it has.
On Work Item. Type datetime. JSON key closed_at.
aow:constraints
Free-text constraints that are not (yet) captured as Policies.
On Intent. Type string-list. JSON key constraints.
aow:contextType
The kind of situational information.
On Context. Type enum. JSON key context_type.
aow:costProfile
Relative cost per invocation.
On Skill. Type enum. JSON key cost_profile.
aow:defaultConfidenceFloor
Confidence below which an Actor operating at this level must escalate, unless a Policy sets a different threshold.
On Assurance Level. Type number. JSON key default_confidence_floor.
aow:dueBy
When the Task must be completed.
On Task. Type datetime. JSON key due_by.
aow:effect
What the Policy does when it applies.
On Policy. Type enum. JSON key effect.
aow:endedAt
When the Action ended.
On Action. Type datetime. JSON key ended_at.
aow:enforcementMode
Whether it blocks before, detects after, or repairs after the fact.
On Guardian. Type enum. JSON key enforcement_mode.
aow:exceptionProtocol
What to do when a Task fails or a Policy blocks progress.
On Plan, Orchestrator. Type text. JSON key exception_protocol.
aow:externalRef
Identifier of the item in its system of record.
On Work Item. Type string. JSON key external_ref.
aow:feedbackType
What the Feedback is about.
On Feedback. Type enum. JSON key feedback_type.
aow:goalStatement
What the Intent is meant to achieve, stated as an outcome rather than a procedure.
On Intent. Type text. JSON key goal_statement.
aow:humanRole
What people do at this level.
On Assurance Level. Type text. JSON key human_role.
aow:idempotencyKey
Key that makes a retried Action safe to repeat.
On Action. Type string. JSON key idempotency_key.
aow:implementationRef
Pointer to the Agent's implementation, model card, or A2A Agent Card.
On Agent. Type uri. JSON key implementation_ref.
aow:inputSchema
JSON Schema (or equivalent) for the Skill's inputs.
On Skill. Type json. JSON key input_schema.
aow:inputs
The inputs passed to the Skill.
On Action. Type json. JSON key inputs.
aow:issueDetected
What went wrong or could be better.
On Feedback. Type text. JSON key issue_detected.
aow:itemType
The kind of work item, for example "insurance-claim" or "purchase-order".
On Work Item. Type string. JSON key item_type.
aow:kpiImpact
The measured change, for example {"cycle_time_hours": {"before": 48, "after": 0.12}}.
On Outcome. Type json. JSON key kpi_impact.
aow:kpiTarget
The measurable target, for example "claims cycle time -20%".
On Objective. Type string. JSON key kpi_target.
aow:latency
Typical or maximum latency, for example "< 1s".
On Skill. Type string. JSON key latency.
aow:level
Position on the AL0 to AL4 scale.
On Assurance Level. Type integer. JSON key level.
aow:measuredAt
When the Outcome was measured.
On Outcome. Type datetime. JSON key measured_at.
aow:measurementWindow
The period the measurement covers.
On Outcome. Type string. JSON key measurement_window.
aow:method
How the value was derived, for example "document-extraction" or "rule".
On Observation. Type string. JSON key method.
aow:name
Display name.
On Actor. Type string. JSON key name.
aow:observedAt
When the Signal occurred.
On Signal, Observation. Type datetime. JSON key observed_at.
aow:observedProperty
What was observed, for example "claim_amount" or "handling_time".
On Observation. Type string. JSON key observed_property.
aow:observedValue
The observed value, in any JSON form.
On Observation. Type json. JSON key observed_value.
aow:openedAt
When the item entered the process.
On Work Item. Type datetime. JSON key opened_at.
aow:organizationalUnit
The team or department the person works in.
On Human Actor. Type string. JSON key organizational_unit.
aow:outputSchema
JSON Schema (or equivalent) for the Skill's outputs.
On Skill. Type json. JSON key output_schema.
aow:overrideProtocol
Who may override the Policy, and how.
On Policy. Type text. JSON key override_protocol.
aow:payloadRef
Where the raw payload is stored, if it is kept.
On Signal. Type uri. JSON key payload_ref.
aow:planningModel
How the Orchestrator plans, for example "rule-based", "LLM-planned", "hybrid".
On Orchestrator. Type string. JSON key planning_model.
aow:priority
Relative business priority.
On Objective, Intent. Type enum. JSON key priority.
aow:rationale
The Actor's stated reason for the Action, recorded for explainability.
On Action. Type text. JSON key rationale.
aow:recommendedAdjustments
Specific changes proposed.
On Feedback. Type string-list. JSON key recommended_adjustments.
aow:resourceConstraints
Limits on concurrency, cost, or time it must respect.
On Orchestrator. Type text. JSON key resource_constraints.
aow:responsibilities
What an Actor in this Role is accountable for.
On Role. Type string-list. JSON key responsibilities.
aow:retentionPolicy
How long entries are kept and when they are purged.
On Memory. Type text. JSON key retention_policy.
aow:reviewProtocol
The human oversight this level requires.
On Assurance Level. Type text. JSON key review_protocol.
aow:riskClass
The class of risk this level is suited to.
On Assurance Level. Type string. JSON key risk_class.
aow:riskProfileChange
How the risk picture changed as a result.
On Outcome. Type text. JSON key risk_profile_change.
aow:ruleLanguage
The language rule_set is written in, for example "ODRL", "Rego", "DMN", "text".
On Policy. Type string. JSON key rule_language.
aow:ruleSet
The rules, or a reference to them.
On Policy. Type text. JSON key rule_set.
aow:scope
The Actors, Skills, or Work Items it watches.
On Guardian. Type text. JSON key scope.
aow:sensitivityLevel
Handling classification of the payload.
On Signal, Context. Type enum. JSON key sensitivity_level.
aow:severity
Consequence of a violation.
On Policy, Feedback. Type enum. JSON key severity.
aow:sideEffects
Whether invoking the Skill changes external state, and whether that can be undone.
On Skill. Type enum. JSON key side_effects.
aow:signalType
The kind of trace.
On Signal. Type enum. JSON key signal_type.
aow:source
The system, sensor, application, or channel that emitted the Signal.
On Signal, Confidence. Type string. JSON key source.
aow:startedAt
When the Action began.
On Action. Type datetime. JSON key started_at.
aow:state
Current business state of the item, in the vocabulary of its system of record.
On Work Item. Type string. JSON key state.
aow:status
Where the Intent is in its life.
On Intent, Plan, Task, Action, Result. Type enum. JSON key status.
aow:storeRef
Where the memory is held.
On Memory. Type uri. JSON key store_ref.
aow:summary
What the Result is, in words.
On Result. Type text. JSON key summary.
aow:timeHorizon
The period over which the Objective is pursued, for example "FY2027".
On Objective. Type string. JSON key time_horizon.
aow:timeToValue
How long it took for the impact to appear.
On Outcome. Type string. JSON key time_to_value.
aow:timestamp
When the Feedback was given.
On Feedback. Type datetime. JSON key timestamp.
aow:triggerConditions
Conditions under which the Policy applies.
On Policy. Type string-list. JSON key trigger_conditions.
aow:validUntil
When the Context should be considered stale.
On Context. Type datetime. JSON key valid_until.
aow:value
Certainty between 0 and 1.
On Confidence. Type number. JSON key value.
aow:variables
The situational values themselves.
On Context. Type json. JSON key variables.
aow:volatility
How much the value is expected to move as evidence arrives.
On Confidence. Type enum. JSON key volatility.
Assurance Levels
aow:AL0 Manual
People: Performs the work. Agents: Observes and prepares; takes no Action that changes a Work Item.
Review: Not applicable; the work is done by people.
aow:AL1 Assisted
People: Decides. Approves every Action that changes a Work Item before it takes effect. Agents: Recommends, drafts, and prepares.
Review: Every state-changing Action carries approved_by.
aow:AL2 Delegated
People: Confirms when confidence is low or a Policy triggers; reviews a sample otherwise. Agents: Decides and acts within Policy.
Review: Escalate below the confidence threshold or on Policy trigger; sampled review otherwise.
aow:AL3 Supervised
People: Reviews by exception, when the Guardian raises an alert, and in periodic audit. Agents: Acts, and must be able to reverse or compensate its Actions.
Review: Guardian-raised exceptions and periodic audit; Skills used must be reversible.
aow:AL4 Autonomous
People: Sets Policy and monitors aggregate Outcomes. Agents: Acts without routine human review; the Guardian can suspend it at any time.
Review: Continuous Guardian monitoring; audit of aggregate Outcomes.