A smallholder's file arrives as it was kept: a statement printed at a branch, a group's delivery ledger written by hand, pages photographed on a phone. Before any judgement can be formed, someone has to establish what the file shows and what it leaves out.
What the reading produces today
On one file, in English or Swahili:
- ●The whole file read, photographed pages included, and each page that could not be read named
- ●The questions the file raises, written down for the officer
- ●An appraisal that says which evidence mattered and why, with underwriting and risk analysis drawn from the same evidence
Why the committee can question it
Every figure shows where it came from. A member who doubts a number can go to the page it was taken from, so the argument is about evidence rather than about whether to trust the tool. Mwikila, LitFin Intelligence's AI Senior Agriculture & Finance Consultant, prepares the analysis. The officer and the committee make the decision: LitFin Intelligence does not approve or decline a loan, and it does not move money.
Training comes with it
The same mind trains the officers who work beside it: lessons on the credit process, role-play on realistic cases, and practice runs of an application from the borrower's side.