I have worked in vehicle finance since 2014. cercit applies that experience to a loan-review system: what information to collect, which rules to check and what an officer needs before deciding a case.
My contribution
I defined the product, credit policy and requirements, then reviewed and tested the build. Claude wrote most of the code.
The loan file
A loan application brings together identity documents, income, credit history and vehicle details. When those disagree or something is missing, an officer needs to see the issue before making a decision.
I defined the first product around salaried new-car buyers. That gave the application a specific document list and a lending policy I could write and review.
What I put on the review screen
The system reads and checks documents, compares income sources and runs lending rules. The officer sees the recommendation beside the reasons, affordability calculations and any failed checks.
I kept the final decision with the officer. They can approve, decline or refer a case, and must record a reason. Missing income or bureau information also goes for review.
How I directed the build
I wrote the requirements and credit policy, set priorities and checked the resulting workflows. Claude handled most implementation, including the website, database and document services.
I also defined how rule changes should happen. A change is drafted, checked for its effect on cases and approved before it becomes active. The system keeps a record of who changed or overrode a decision.
Try it with synthetic cases
You can try the customer and staff journeys using synthetic cases. The public repository includes the requirements, architecture and test documentation.
The demo shows the workflow. To measure its effect on approval time or credit quality, the next step would be a lender pilot with agreed measures and representative cases.