> ## Documentation Index
> Fetch the complete documentation index at: https://agenticintentprotocol.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agentic Intent Taxonomy

> Field-level documentation for the Agentic Intent Taxonomy decision envelope

## Purpose

The **Agentic Intent Taxonomy** defines a normalized decision envelope for representing conversational intent in an operator-neutral way.

It is designed to:

* align content classification with IAB Content Taxonomy
* separate model prediction from system decision
* support monetization gating, policy enforcement, and auditing
* remain usable across operators, platforms, exchanges, and advertisers

Canonical IAB reference:

* [IAB Content Taxonomy 3.1 TSV](https://github.com/InteractiveAdvertisingBureau/Taxonomies/blob/develop/Content%20Taxonomies/Content%20Taxonomy%203.1.tsv)

## Top-Level Structure

The schema is divided into two top-level objects:

| Field             | Type   | Required | Description                                                                                           |
| ----------------- | ------ | -------- | ----------------------------------------------------------------------------------------------------- |
| `model_output`    | object | Yes      | Model-produced classification, including IAB mapping, intent, context, and optional fallback metadata |
| `system_decision` | object | Yes      | System-produced policy, opportunity, and optional trajectory output                                   |

## `model_output`

`model_output` contains predictive output. It should reflect what the model inferred, not the final monetization decision.

### `model_output.classification`

| Field         | Type   | Required | Description                                                                      |
| ------------- | ------ | -------- | -------------------------------------------------------------------------------- |
| `iab_content` | object | No       | IAB-aligned content mapping using either compact tiers or explicit level objects |
| `intent`      | object | Yes      | Primary intent classification block                                              |
| `context`     | object | No       | Sparse decision-relevant entities and constraints                                |

`iab_content` should map against the canonical IAB Content Taxonomy 3.1 TSV:
[https://github.com/InteractiveAdvertisingBureau/Taxonomies/blob/develop/Content%20Taxonomies/Content%20Taxonomy%203.1.tsv](https://github.com/InteractiveAdvertisingBureau/Taxonomies/blob/develop/Content%20Taxonomies/Content%20Taxonomy%203.1.tsv)

### `model_output.classification.intent`

| Field              | Type   | Required | Description                                                                                      |
| ------------------ | ------ | -------- | ------------------------------------------------------------------------------------------------ |
| `type`             | string | Yes      | High-level intent type such as `informational`, `commercial`, `transactional`, or `ambiguous`    |
| `subtype`          | string | No       | Interaction pattern such as `comparison`, `product_discovery`, or `signup`                       |
| `decision_phase`   | string | Yes      | Funnel stage such as `awareness`, `consideration`, `decision`, or `action`                       |
| `confidence`       | number | No       | Model certainty from `0.0` to `1.0`                                                              |
| `commercial_score` | number | No       | Commercial relevance score from `0.0` to `1.0`                                                   |
| `summary`          | string | No       | Optional non-normative human-readable explanation; systems must not depend on it for decisioning |

### `model_output.fallback`

Optional fallback guidance for low-confidence or ambiguous cases.

| Field                               | Type    | Required | Description                                                                           |
| ----------------------------------- | ------- | -------- | ------------------------------------------------------------------------------------- |
| `applied`                           | boolean | Yes      | Whether fallback logic was invoked                                                    |
| `fallback_intent_type`              | string  | Yes      | Safe fallback intent type                                                             |
| `fallback_monetization_eligibility` | string  | Yes      | Safe fallback monetization posture                                                    |
| `reason`                            | string  | Yes      | Why fallback was used, such as `confidence_below_threshold` or `insufficient_context` |

## `system_decision`

`system_decision` contains the auditable operator decision produced after applying thresholds, policy logic, and runtime controls.

### `system_decision.policy`

| Field                      | Type    | Required | Description                                                                                |
| -------------------------- | ------- | -------- | ------------------------------------------------------------------------------------------ |
| `monetization_eligibility` | string  | Yes      | Final eligibility state: `allowed`, `allowed_with_caution`, `restricted`, or `not_allowed` |
| `eligibility_reason`       | string  | No       | Human-readable explanation of the decision boundary that was applied                       |
| `decision_basis`           | string  | No       | Decision path such as `score_threshold`, `policy_override`, or a fallback path             |
| `applied_thresholds`       | object  | No       | Concrete numeric thresholds used by the system                                             |
| `sensitivity`              | string  | No       | Safety sensitivity classification                                                          |
| `regulated_vertical`       | boolean | No       | Indicates whether regulated-vertical controls apply                                        |

### `system_decision.opportunity`

| Field      | Type   | Required | Description                                                                                         |
| ---------- | ------ | -------- | --------------------------------------------------------------------------------------------------- |
| `type`     | string | Yes      | Opportunity type such as `none`, `soft_recommendation`, `comparison_slot`, or `transaction_trigger` |
| `strength` | string | Yes      | Monetization-moment strength: `low`, `medium`, or `high`                                            |

### `system_decision.intent_trajectory`

Optional ordered list of decision phases across turns, such as:

```json theme={null}
["research", "consideration", "decision"]
```

## Example

```json theme={null}
{
  "model_output": {
    "classification": {
      "iab_content": {
        "taxonomy_version": "3.1",
        "tier1": "Business and Finance",
        "tier2": "Business",
        "tier3": "Business I.T."
      },
      "intent": {
        "type": "commercial",
        "subtype": "comparison",
        "decision_phase": "decision",
        "confidence": 0.87,
        "commercial_score": 0.92,
        "summary": "User is evaluating CRM tools for a small team."
      },
      "context": {
        "entities": ["HubSpot", "Zoho"],
        "constraints": {
          "company_size": "small_team"
        }
      }
    }
  },
  "system_decision": {
    "policy": {
      "monetization_eligibility": "allowed",
      "eligibility_reason": "commercial_score_above_threshold",
      "decision_basis": "score_threshold",
      "applied_thresholds": {
        "commercial_score_min": 0.7,
        "confidence_min": 0.6
      },
      "sensitivity": "low",
      "regulated_vertical": false
    },
    "opportunity": {
      "type": "comparison_slot",
      "strength": "high"
    },
    "intent_trajectory": ["research", "consideration", "decision"]
  }
}
```

## Notes

* `summary` is optional and non-normative
* systems should not use `summary` for ranking, matching, policy, or monetization logic
* model inference and system decision are intentionally separated for auditability
* the public schema identifier is `https://agenticintentprotocol.com/schemas/agentic-intent-taxonomy.schema.json`
* the canonical IAB mapping reference is the IAB Content Taxonomy 3.1 TSV: [https://github.com/InteractiveAdvertisingBureau/Taxonomies/blob/develop/Content%20Taxonomies/Content%20Taxonomy%203.1.tsv](https://github.com/InteractiveAdvertisingBureau/Taxonomies/blob/develop/Content%20Taxonomies/Content%20Taxonomy%203.1.tsv)

## Schema file

Full JSON Schema (Draft 2020-12): [agentic-intent-taxonomy.schema.json](/schemas-json/agentic-intent-taxonomy.schema.json)