Decisions

What it is

The record of process decisions. Each row is one question raised by the business — a bottleneck someone is stuck on, an assumption an employee is about to act on, a rule waiting to be approved. A row captures the full picture: who raised it and in which area (identity, topic), what is actually happening (situation), the assumption and its risk (assumption), how it is classified (reversibility, financial and risk levels), whether it matches a recurring pattern, the human response once someone decides, the rule it turns into, the documents it touches, its visibility, and where it sits in the workflow. Agents can raise a decision and propose an answer, but the response and the resulting rule are filled in by a person — see the note under Fields.

How it gets data

Ingest-enabled — external systems and agents push rows through the Ingest API:

POST /api/ingest/decisions
Authorization: Bearer pk_...
Content-Type: application/json

One JSON record per request — one row per question.
Idempotent upsert on (tenant_id, external_id) — re-sending the same external_id updates the row, never duplicates it.
tenant_id / workspace_id come from the API key context, never from the body.

Fields

The question (Decision V2)

The slim core every decision needs. Agents and people fill these first; the long-form sections below are optional detail and are hidden from the default columns.

Field	Type	Description
`decision_question`	text	The one question a human has to answer
`recommendation`	text	The proposed answer, in plain language
`options`	json	Array of options, e.g. `[{"label": "...", "pros": "...", "cons": "..."}]` — agents must send a real JSON array, not a string
`process_id`	uuid	The Process this decision belongs to (late-bound; may be assigned later by a curator or person)
`agent_id` / `run_id`	uuid	Provenance — stamped server-side when an agent creates the row, never accepted from the payload

Evidence for a decision lives in Decision Signal Links: the Company Signals that triggered, support, or contradict it. In the app they render as the Evidence tab.

Identity

Field	Type	Description
`decision_no`	text	Human-readable decision number
`title`	text	Short title of the decision
`external_id`	text	Source id — the upsert key
`department_id`	uuid	Department the decision belongs to
`role_label`	text	Role of the person who raised it
`submitted_by_user_id`	uuid	User who submitted the decision
`interview_mode`	text	How it was captured (e.g. interview, quick)

Topic

Field	Type	Description
`topic_type`	text	Kind of decision (e.g. process, policy, exception)
`process_area`	text	Which process area it concerns
`affected_departments`	text	Departments affected by the decision

Situation

Field	Type	Description
`current_situation`	text	What is happening now
`problem_or_bottleneck`	text	The problem or bottleneck being hit
`waiting_for`	text	What the work is waiting on
`blocked_work_count`	integer	How much work is blocked by this
`waiting_since`	datetime	Since when it has been waiting
`decision_needed_by`	datetime	Deadline for a decision

Assumption

Field	Type	Description
`employee_assumption`	text	What the employee assumes the answer is
`reason_for_assumption`	text	Why they assume it
`risk_if_wrong`	text	What happens if the assumption is wrong
`confidence`	text	How confident they are

Classification

Field	Type	Description
`reversible_decision`	boolean	Whether the decision can be undone
`could_become_permanent_rule`	boolean	Whether it might become a standing rule
`proposed_decision_level`	text	Proposed level at which it should be decided
`financial_impact`	number	Estimated financial impact
`financial_currency`	text	Currency of the financial impact
`risk_level`	text	Assessed risk level
`impact_level`	text	Assessed impact level
`decision_owner_role`	text	Role that owns the decision

Pattern

Field	Type	Description
`pattern_key`	text	Key grouping similar decisions
`similar_case_count`	integer	How many similar cases have been seen
`pattern_summary`	text	Summary of the recurring pattern
`recommended_permanent_rule`	text	Rule the pattern suggests adopting

Response (human-decided)

Field	Type	Description
`decision_response`	text	The answer given
`decision_modification`	text	Any modification to the proposed answer
`decision_reason`	text	Reasoning behind the answer
`answered_by_role`	text	Role of the person who answered
`answered_by_type`	text	Whether answered by a human or escalation
`answered_at`	datetime	When it was answered
`final_decision`	text	The final decision text
`final_decision_level`	text	Level at which it was finally decided

Rule (human-decided)

Field	Type	Description
`conditions`	text	Conditions under which the rule applies
`escalation_conditions`	text	When to escalate instead of applying the rule
`financial_limit`	number	Financial ceiling the rule permits
`risk_limit`	text	Risk ceiling the rule permits
`effective_from`	date	When the rule takes effect
`review_date`	date	When the rule should be reviewed
`rule_version`	text	Version of the rule

Documents

Field	Type	Description
`related_sop`	text	Related SOP
`related_skill`	text	Related agent skill
`related_strategy`	text	Related Strategy document
`document_change_required`	text	What document needs to change as a result
`published_document`	text	The document published from this decision
`published_version`	text	Version of the published document
`published_at`	datetime	When it was published
`published_by_role`	text	Role that published it

Visibility

Field	Type	Description
`scope`	text	Who can see the decision (the “Who sees this?” scope — see Data Views)

Workflow

Field	Type	Description
`status`	text	Where the decision is in its lifecycle
`priority`	text	Priority of the decision
`notification_required`	boolean	Whether a notification must go out
`notification_reason`	text	Why a notification is required
`closed_at`	datetime	When the decision was closed
`cause`	text	Root cause, for process issues
`action`	text	Agreed action
`due_date`	date	When the action is due

Agents can raise a decision and propose an answer — they fill the situation, assumption, classification, and pattern fields. The Response, Rule, and published fields are human-decided: an agent may recommend, but a person fills them in. metadata (json) is accepted on ingest but not surfaced as a default column.

Like every workspace table, Decisions also carries the system columns id, tenant_id, workspace_id (tenant isolation), created_at, and updated_at. They are managed automatically.

Query it

{
  "method": "tools/call",
  "params": {
    "name": "data_query",
    "arguments": {
      "view_slug": "decisions",
      "filters": { "status": "open" },
      "select_fields": ["decision_no", "title", "decision_question", "recommendation", "priority", "decision_needed_by"],
      "sort_by": "decision_needed_by",
      "limit": 50
    }
  }
}

Getting Started

Vertical Intelligence

AI Operating Guide

Resources

Sales & Commerce

Procurement

Marketing & Outreach

Governance

What it is

How it gets data

Fields

The question (Decision V2)

Identity

Topic

Situation

Assumption

Classification

Pattern

Response (human-decided)

Rule (human-decided)

Documents

Visibility

Workflow

Query it

​What it is

​How it gets data

​Fields

​The question (Decision V2)

​Identity

​Topic

​Situation

​Assumption

​Classification

​Pattern

​Response (human-decided)

​Rule (human-decided)

​Documents

​Visibility

​Workflow

​Query it

What it is

How it gets data

Fields

The question (Decision V2)

Identity

Topic

Situation

Assumption

Classification

Pattern

Response (human-decided)

Rule (human-decided)

Documents

Visibility

Workflow

Query it