AI agent training for underwriters in consumer lending in South Africa

Underwriters at South African consumer lenders must be trained to supervise agents that run affordability calculations, test payslips against bank data and check debt review status, while declines, exceptions and disbursements stay with them. Training covers verdicts, provenance and held approvals, plus POPIA's limits on automated decisions and the national affordability assessment rules.

What does a consumer lending underwriter in South Africa do?

Unsecured personal loans, store credit, credit cards and vehicle finance make up most consumer lending. Many borrowers already carry several obligations, so the affordability assessment, which compares verified income with existing debt and living expenses, decides most files. Lenders must take reasonable steps to assess whether the borrower can afford the credit, and lending without doing so can make the agreement reckless and unenforceable.

Underwriters see payslips, bank statements and credit bureau reports. Payslip fraud is common: edited figures, fictitious employers, or salary credits that do not match. Some applicants are under debt review, which bars new credit, and some have salaries already subject to court-ordered deductions. Underwriters approve, decline or refer, and every decision needs to be defensible.

Where do agents help, and how far may they act?

1. Payslip and bank statement reconciliation, depth 1 to 2. The agent compares the payslip's net pay with salary credits in bank statements obtained with consent, and flags differences in amount, date or employer name. Bank data received directly is signed; an uploaded payslip is declared.
2. Affordability calculation, depth 2. The agent applies the lender's affordability model, using verified income, bureau obligations and the expense norms the lender applies, and drafts the assessment with every figure sourced.
3. Debt review and court order checks, depth 1. The agent reads the bureau report for debt review flags and existing deductions and stops the file where one is present.
4. Recommendation drafting, depth 2. The agent drafts a recommendation with reasons for the underwriter.
5. Disbursement, depth 4. Held for a named approver above the lender's limit; a payout account that is not in the borrower's name is held every time.

What must underwriters learn, and what stays with them?

Begin with what underwriters should learn before working with agents. For a South African lender, the emphasis is on the affordability file. An underwriter must be able to trace every income and expense figure to its source, see where the agent used declared evidence, and know when a mismatch between payslip and bank data is a timing issue and when it is fraud.

Underwriters also need to read holds correctly. A file stopped for debt review is the control working; releasing it would create reckless credit. A file held because the agent's certificate does not cover vehicle finance is a scope issue, not a judgement about the borrower.

Held items need timely review, especially at month end when applications peak; a reviewer rota for the lending team keeps them moving.

Underwriters keep:

• every decline and exception;
• the final affordability judgement where figures are borderline;
• fraud conclusions on payslips and identities;
• decisions about borrowers showing signs of over-indebtedness.

Which South African rules shape oversight and competence?

National credit legislation requires lenders to conduct an affordability assessment before granting credit and treats lending without a proper assessment as reckless, with serious consequences for enforceability. A dedicated credit regulator oversees compliance. An agent can do the calculations and evidence gathering, but the lender must be able to show the assessment was properly made, and a competent person should own it.

The Protection of Personal Information Act (POPIA), enforced by the Information Regulator, governs borrower data. Its conditions for lawful processing apply to every agent, and it protects people against decisions based solely on automated processing that substantially affect them, subject to specific exceptions and safeguards. A credit refusal is such a decision; keep a person in it. See POPIA and training staff who use AI agents for the training implications.

Banks answer to the Prudential Authority for safety and soundness, and the FSCA oversees market conduct and fair treatment of customers. For the training picture across sectors, see what AI agent training organisations in South Africa need.

This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.

A 90-day training plan

Credit managers and collections staff need different skills again; AI agent skills by role compares them.

Where the Academy fits

Pryme Intelligence Academy offers a free readiness assessment for lending teams of up to 25 people: the team takes the Foundations test set in lending, and the report shows scores by concept, the gaps and the first agent to start with. Underwriters who approve held disbursements then take the Operator track and become Certified Pryme Operator.