Agentic Ontology of work

# A shared language for the agentic enterprise

The Agentic Ontology of Work (AOW) is Skan AI's open semantic framework for describing, governing, and scaling AI agents across enterprise operations.

![Manish and Avinash](https://cdn.sanity.io/images/3xg3qj5k/production/2d1917b7759d7c075d664cc815e2967da92dce21-7459x4972.jpg?q=80&fit=max&auto=format)

Less confusion, more shared language

Different teams use different terms for the same concepts. AOW creates a single, open vocabulary for intelligent work, so engineering, operations and AI teams can describe what they're building without losing each other in translation.

1.  Less confusion, more shared language
    
    Different teams use different terms for the same concepts. AOW creates a single, open vocabulary for intelligent work, so engineering, operations and AI teams can describe what they're building without losing each other in translation.
    
2.  Govern AI at scale
    
    AOW makes policy, assurance, explainability and accountability part of how autonomous systems are built. The result is agents that work safely at enterprise scale, with the governance evidence ready before it's asked for.
    
3.  Enable interoperability across your stack
    
    Agents, workflows, BPM systems, RPA and LLMs all need to speak to each other. AOW connects them through one semantic framework, so the agentic estate can grow without locking the enterprise into any single platform.
    
4.  From SOA to agentic AI
    
    Service-oriented architecture gave enterprises a shared vocabulary for services and contracts. The agentic era needs a richer framework, one that describes reasoning, autonomy and governance, not just APIs and endpoints. AOW is that framework
    

-   [Agentic Ontology Model](/agentic-ontology#agentic-ontology-model)
-   [Vocabulary](/agentic-ontology#vocabulary)
-   [Whitepaper](/agentic-ontology#whitepaper)
-   [Related products](/agentic-ontology#related-products)

Agentic Ontology Model

## The semantic graph
of intelligent work

The model shows how the system works as a whole: how objectives, agents, policies, outcomes and feedback connect to enable intelligent work across the enterprise.

-   ### Objective
    
    The business goal that everything else in the model exists to serve.
    
    ## Objective
    
    The business goal that everything else in the model exists to serve.
    
    **Definition:** A business-level goal, expressed in strategic or operational terms and typically tied to a KPI, SLA, or regulatory requirement.
    
    **Relationships:** Spawns one or more Intents. Evaluated through Outcomes.
    
    **Governance Role:** Anchors agentic activity to enterprise value, so autonomous work can always be traced back to a business reason.
    
-   ### Intent
    
    The machine-readable version of that goal, specific enough for an agent to act on.
    
    ## Intent
    
    The machine-readable version of that goal, specific enough for an agent to act on.
    
    **Definition**: A structured, machine-interpretable goal derived from an Objective and specific enough for an Agent or Orchestrator to act on. Where the Objective is the why, the Intent is the what.
    
    **Relationships:** Shaped by Context, constrained by Policy, decomposed by the Orchestrator, and fulfilled by one or more Agents.
    
    **Governance Role:** Carries the assurance requirement that determines how much oversight its execution needs.
    
-   ### Context
    
    The real-time detail that tells an Intent what's actually going on.
    
    ## Context
    
    The real-time detail that tells an Intent what's actually going on.
    
    **Definition:** The structured situational information, customer data, system state, environment, history, that gives an Intent its relevance.
    
    **Relationships:** Informs Intent and is drawn from Memory. Updated Context produces sharper future Intents.
    
    **Governance Role:** Grounds decisions in real conditions rather than static assumptions, supporting explainability.
    
-   ### Orchestrator
    
    The planner that turns an Intent into a sequence of tasks agents can run.
    
    ## Orchestrator
    
    The planner that turns an Intent into a sequence of tasks agents can run.
    
    **Definition:** The coordinating entity that decomposes an Intent into executable tasks, sequences them, and assigns them to Agents.
    
    **Relationships:** Receives Intent, consults Policy, and delegates to Agent.
    
    **Governance Role:** Keeps execution inside policy boundaries by only planning what governance allows.
    
-   ### Agent
    
    The actor that does the work, one Skill at a time.
    
    ## Agent
    
    The actor that does the work, one Skill at a time.
    
    **Definition:** A policy-aware actor capable of interpreting Intents and acting on them using Skills.
    
    **Relationships:** Invokes Skills, produces Results, and writes to Memory. Operates within limits set by Policy.
    
    **Governance Role:** Must respect its assigned autonomy level and required Assurance Level, and is accountable for every action it takes.
    
-   ### Skill
    
    The specific capability an Agent calls on to get a task done.
    
    ## Skill
    
    The specific capability an Agent calls on to get a task done.
    
    **Definition:** A reusable, atomic capability that an Agent can call to perform a specific piece of work.
    
    **Relationships:** Invoked by Agent. Produces a Result.
    
    **Governance Role:** Executes only within the constraints Policy sets, and its output is subject to Assurance verification before it can be trusted.
    
-   ### Result
    
    The immediate, atomic output of a Skill doing its job.
    
    ## Result
    
    The immediate, atomic output of a Skill doing its job.
    
    **Definition:** The immediate effect or atomic output produced when a Skill executes.
    
    **Relationships:** Aggregates into Outcome. Annotated with a Confidence measure and supporting artifacts (documents, logs, state changes).
    
    **Governance Role:** Checked against the required Assurance Level before it can count toward an Outcome.
    
-   ### Outcome
    
    The business impact once individual results add up.
    
    ## Outcome
    
    The business impact once individual results add up.
    
    **Definition:** The aggregated business impact of one or more Results, measured in terms like cycle time, error rate, or SLA performance.
    
    **Relationships:** Evaluated against the original Objective. Generates Feedback.
    
    **Governance Role:** The point where agentic work gets measured against the business case that justified it in the first place.
    
-   ### Feedback
    
    What the system learns from what just happened, fed back in.
    
    ## Feedback
    
    What the system learns from what just happened, fed back in.
    
    **Definition:** Structured information generated from execution and used to improve future behavior.
    
    **Relationships:** Updates Policy and Memory, and refines how future Intents are decomposed.
    
    **Governance Role:** Closes the loop between what happened and what governance allows next time.
    
-   ### Memory
    
    Where that learning gets stored so the next decision is better informed.
    
    ## Memory
    
    Where that learning gets stored so the next decision is better informed.
    
    **Definition:** Persistent storage of knowledge gained from experience, spanning short-lived Operational Memory and a longer-term Knowledge Base.
    
    **Relationships:** Enriches Context, is updated by Feedback, and maintains Provenance for everything it stores.
    
    **Governance Role:** Preserves an auditable record of past decisions and results, supporting both learning and compliance review.
    
-   ### Guardian
    
    The check that keeps every other entity accountable.
    
    ## Guardian
    
    The check that keeps every other entity accountable.
    
    **Definition:** The oversight mechanism that enforces governance, policy adherence, explainability, and auditability across the system.
    
    **Relationships:** Connects to every entity in the model. Checks Agents against Policy, verifies Assurance Levels against Confidence, and confirms Provenance.
    
    **Governance Role:** The primary enforcement point. Nothing in the ontology is treated as trustworthy until Guardian has validated it.
    

Vocabulary

## The vocabulary of intelligent work

Perception

**Telemetry:** The granular, timestamped record of individual events and state changes generated as work happens, the raw material the Perception layer captures first.  
  
**Signals:** The full range of captured inputs, events, states, documents, logs, and interactions, that make up the ground truth of what is happening across the enterprise.  
  
**Observations:** The environmental and system-level detail (what changed, where, and around what) that gives Signals their real-world setting.  
  
**Context:** The structured situational information that gives relevance to an Intent.

Cognition

**Objective:** A business-level goal expressed in strategic or operational terms.

**Intent:** A structured, actionable goal derived from an Objective and interpretable by Agents or Orchestrators.

**Policy:** Declarative constraints that define what behavior is permissible, required, or restricted.

**Orchestrator:** A coordinating entity that decomposes Intents into executable tasks and assigns them to Agents.  

Execution

**Agent:** A policy-aware actor that interprets Intents and acts on them using Skills.

**Skill:** A reusable capability that an Agent can call to perform a specific piece of work.

**Result:** The immediate effect or atomic output produced when a Skill executes.

Assurance

**Confidence:** A model or rule-derived measure of certainty about a perception, decision, or action.

**Assurance Level:** A governance-required threshold that determines how much autonomy an action gets versus how much human oversight it needs.

**Feedback:** Structured information used to improve future behavior.

**Memory:** Persistent storage of knowledge from experience, covering both short-lived Operational Memory and a long-term Knowledge Base.

**Guardian:** The oversight mechanism that enforces governance, policy adherence, explainability, and auditability.

**Provenance:** Metadata that captures the lineage, origin, and evolution of every entity in the ontology, so any decision can be reconstructed end to end.

Whitepaper

## Read the framework in full

The complete AOW whitepaper covers the model, the vocabulary, the implementation patterns, and the governance posture in detail, written for enterprise architects and AI transformation leads.

[Download the whitepaper](/whitepapers/agentic-ontology-of-work)

[![AI Agents Needed a Common Language. So We Built One.](https://cdn.sanity.io/images/3xg3qj5k/production/4f4a0c6d85fd49f816af4ccb532d7327540fec9d-1308x873.png?q=80&fit=max&auto=format)

](/whitepapers/agentic-ontology-of-work)

## AOW in production,
  
across the Skan AI platform

Blueprint

[

![Your AI masterplan](https://cdn.sanity.io/images/3xg3qj5k/production/33ba832e4a26326626c548e23e88a16d205e6896-536x464.svg)

](/blueprint)

Your AI masterplan

Recommendations, roadmaps and ROI.

[Learn more](/blueprint)

See how work really happens

Maps, metrics and opportunities, built from real work.

[Learn more](/process-intelligence)

See where engineering value is built

Signals, patterns and productivity, across the full developer day.

[Learn more](/engineering-intelligence)

Deploy agents grounded in your reality

Decisions, actions and outcomes, grounded in operational context.

[Learn more](/agents)

Your AI masterplan

Recommendations, roadmaps and ROI.

[Learn more](/blueprint)