What happens when an intake agent finds a loan file incomplete?
A loan application at Aster Financial arrives with a mismatched income figure and an unreadable identity document, and the agent must hold it rather than route it onward.
At Aster Financial, application APP-2187 for a personal loan arrives through the broker portal at 4:50 pm on a Friday. The intake agent reads the submission and finds the declared income is higher than the figure that the attached payslip annualises to. The identity document is a driver licence photo with the expiry date cut off.
The agent checks the completeness policy POL-INT-04, which requires a legible identity document and one income proof that matches the declared income within a stated tolerance. Neither condition is met. Instead of routing to the credit officer queue, the agent changes the status of APP-2187 from RECEIVED to PENDING_DOCS, writes a reason code for each failed check, and sends the broker a request listing the two items needed.
The consequence shows on Monday. Because the case was not routed as complete, the credit officer queue contains only files ready for assessment. Had the agent routed it anyway, the credit officer would have spent time rediscovering the same gaps, and the broker would have received a vague query days later.
**What it teaches:** intake quality depends on state changes that are correct, explained and reversible. An evaluation needs the final status, the reason codes, the message sent and the absence of a wrong routing, not just a plausible summary.
Correct intake means explained, reversible state changes, so evaluation should check status, reason codes and messages rather than a plausible summary.
- Audience:
- Agent developers
- Content type:
- Scenario
- Buyer question:
- What it is and why it matters
Content release: r-a220eb6e099b243b
Related questions
- Which signs show an intake process is worth modelling for agents?
- Which systems and links does an intake agent work across?
- What does application intake look like inside a real business?
- Why is intake a good test for AI agents?
- What does a single intake decision look like when an agent has to change state?
- How does intake map onto environment entities, links and state transitions?