completion-vs-acceptance

Completion vs acceptance gap analysis measures the difference between an agent's claimed finished tasks and downstream acceptance rates using structured event logs. It computes per-workflow-type metrics to identify under-trusting or overconfident autonomous behaviors.

A trust quadrant classification maps workflow types onto a 2x2 matrix using completion and acceptance rates. It categorizes agent outputs to generate autonomy recommendations and highlight where mitigations are needed for misaligned trust levels.

To calculate completion and acceptance rates, provide a structured event log containing task_id, workflow_type, status, and accepted fields. The analysis processes this data over a configurable time window with a minimum sample size to produce per-type rates.

Workflow observability metrics apply across various workflow types including code-review, email-draft, and db-migration. The analysis produces per-type completion and acceptance rates, quadrant classification, and an autonomy signal for each distinct category.

Generating an autonomy signal requires a structured event log with fields like task_id, workflow_type, status, and accepted or a proxy field. You must also configure a time window and minimum sample size to ensure sufficient data for reliable trust quadrant classification.

High completion with low downstream acceptance indicates an overconfident agent within the trust quadrant framework. This gap analysis identifies the specific workflow types causing the mismatch and recommends targeted mitigations to realign autonomy levels.