# Human Agent Workflow

### The Pull-Based Assignment Model

The Review Platform uses a pull-based assignment model rather than automatically pushing work to agents. This design choice addresses several common operational challenges that occur with automatic assignment systems.

**Why pull-based assignment matters:**

**Prevents cherry-picking**: When reviewers can see and select from a list of available reviews, they naturally gravitate toward easier or more familiar cases, leaving complex reviews sitting in the queue. With pull-based assignment, the system assigns the highest-priority unassigned review automatically when an agent requests work.

**Improves work-in-progress visibility**: Push-based systems often assign work based on theoretical capacity ("Agent has fewer than 10 cases assigned"), but this doesn't reflect actual work-in-progress. An agent might be deep in research on one complex case while having 5 others "assigned" but not actively worked. Pull-based assignment ensures agents only receive new work when they explicitly request it.

**Maintains fairness**: The system ensures equitable distribution because no agent can claim easier work ahead of others. Everyone receives the next highest-priority review in their queue when they're ready.

**How it works:**

1. **Request Next Review**: Reviewer clicks "Pull Next Review" button in their queue
2. **Automatic Assignment**: System assigns the **highest-priority** unassigned review from that queue
3. **Complete or Unassign**: Agent must finish the review or release it back to the queue before requesting another

This workflow ensures that urgent reviews don't sit unassigned because agents are working on easier cases or have work assigned but not actively in progress.

### Self-Assignment for Special Cases

While pull-based assignment handles the majority of reviews, the platform also supports self-assignment for specific scenarios:

* **Follow-up Reviews**: An agent who reviewed a customer previously can self-assign related reviews
* **Specialized Expertise**: Subject matter experts can claim reviews that require specific knowledge
* **Escalated Cases**: Supervisors can self-assign cases escalated from their team