What does a sufficiency decision look like on a real loan application?
At Aster Financial, an agent checks application APP-2187 against its requirement set, finds three gaps, withholds the file from the credit officer and sends one precise request.
At Aster Financial, an intake analyst opens APP-2187, a personal loan application with status `received`. The bundle holds a signed application form, two payslips, a bank statement and a driver licence image. The requirement set for employed applicants asks for three recent payslips and a statement covering the most recent period.
The agent compares the bundle to the requirement set. Only two payslips are present, and the bank statement ends five weeks before the application date. The application also lists a second income source with no supporting document.
The agent does not forward the file to the credit officer. It sets the status to `awaiting-information`, records three gaps against the requirement IDs, and sends one request naming each item, the period needed and the reason. The credit officer's queue stays free of an incomplete file.
Two days later the applicant supplies the third payslip and a current statement but ignores the second income source. The agent re-evaluates, closes two gaps, keeps one open, and asks again, this time only for that item.
**What it teaches:** sufficiency changes as documents arrive. A good agent tracks gaps individually, avoids repeating requests already satisfied, and treats a partial reply as progress rather than as completion or failure.
Sufficiency is a state that changes as documents arrive, so evaluation needs the whole sequence of requests and replies, not one snapshot.
- Audience:
- Agent developers
- Content type:
- Scenario
- Buyer question:
- What it is and why it matters
Content release: r-a220eb6e099b243b
Related questions
- What does it mean for an agent to decide whether the supplied documents are enough?
- How do I tell whether my process suits sufficiency-focused training or evaluation?
- Where does the sufficiency decision sit in an agent loop, and which tool actions does it need?
- Why not handle this with a fixed checklist or rules engine instead of an agent?
- How does a sufficiency decision map onto environment state and tool actions?
- What makes a sufficiency decision worth putting in a simulated environment?