What might a reseller pilot built on the Aster Financial sample look like?

A marketplace's catalogue manager wants to list a receivables product for a prospective buyer's ML lead, who is testing an agent for cash application at a lender. They scope a pilot around Aster Financial, MirrorFirm's sample organisation, covering loan repayments and fee payments.

The pilot request, tracked as `PILOT-0093`, asks for three things: a dataset of generated repayment bank lines linked to loan accounts, a task pack of 'apply or hold' questions with labels in a separate file, and a short note on how the material was grounded and which disclosure checks to run.

When the sample arrives, the catalogue manager reads cases with a credit operations reviewer. Case `CASE-0207` has a repayment from a joint holder under a different name and no reference. The expected outcome is a hold with a reason code, not a match. The ML lead runs a baseline and a candidate agent against the task pack, using the scorer supplied.

**State change:** `PILOT-0093` moves from requested to sample reviewed, with noted defects and an agreed extension of exception types before any onward listing.

**What it teaches:** a pilot ties the commercial step to evidence. Both the reseller and the buyer read cases, score a baseline, and record what must change before the product is listed.