Nigerian consumer lending operations can give AI agents the checking work: matching BVN and NIN records, verifying salary credits, reconciling failed repayments and drafting compliant reminders. Credit decisions, recovery escalation and anything touching a borrower's reputation stay with people. The CBN, the Nigeria Data Protection Act 2023 and consumer protection rules for digital lenders apply.
What takes up an operations leader's time in Nigerian consumer lending?
Identity first. Every application has to tie a bank verification number, a national identification number, a phone number and a bank account to one person, and the records rarely agree perfectly. Middle names drop out, dates of birth differ by a day, and a phone line is registered to a relative. Each mismatch needs someone to decide whether it is a clerical difference or a sign of fraud.
Income second. Salary-backed loans depend on proving that a salary lands in the borrower's account each month from a named employer. Operations reads statements, checks employer names against an approved list, and watches for accounts that receive a salary-sized transfer once and then go quiet.
Repayments third. Card-on-file debits fail, direct debits are cancelled, and transfers arrive with references that match nothing. On top sits the reputational legacy of the sector: borrowers and regulators are alert to aggressive recovery, and every contact a team makes is now examined.
Where can AI agents help?
- Identity record matching (depth 2). The agent compares BVN, NIN, phone and account details, applies the lender's tolerance rules for name order and spelling, and drafts a pass, a query or a fraud referral. An analyst clears every query and every referral.
- Salary verification (depth 2). The agent reads the statement, identifies recurring credits from the declared employer, checks the employer against the approved list and drafts the verified income figure with the transactions it relied on. A credit officer uses it.
- Repayment reconciliation (depth 3). The agent matches incoming transfers to loan accounts by amount, sender name and reference, posts confident matches within limits and holds the rest for a person.
- Reminder drafting (depth 2). The agent drafts reminders from approved templates, addressed only to the borrower, at permitted times. A collections officer sends or approves the batch.
- Settlement quotes (depth 3). The agent calculates the payoff figure and issues it on the approved template.
Recovery from other accounts deserves care. Banks can, under the CBN's arrangements, recover a defaulted loan from a borrower's accounts at other banks when the borrower agreed to it at origination. An agent can prepare the case file: the consent record, the default history, the amounts. Triggering recovery is a money movement that is hard to undo, and a person should start it every time.
What stays with people, and why?
Credit decisions stay with a credit officer. The agent's verified income and identity file make that decision faster and better documented, but the lender answers for whom it lends to and on what terms, and that accountability should rest with a named person.
Recovery escalation stays human. Nigerian rules for digital lenders, enforced by the federal consumer protection regulator, prohibit debt-shaming: contacting a borrower's phone contacts, publishing their details, or threatening messages. An agent has no business deciding who to contact or what to say beyond an approved reminder. It should be designed so that it cannot read a borrower's contact list at all.
Complaints stay human. A borrower who says a loan was disbursed without consent, or that they repaid and were still chased, is raising something that could be fraud, a system error or misconduct, and each needs a person who can investigate and correct.
Which rules apply in Nigeria?
The Central Bank of Nigeria licenses and supervises banks, microfinance banks and finance companies, and sets rules on KYC, BVN use, consumer protection and recovery arrangements. Lenders that are not CBN-licensed but lend digitally are registered and supervised for conduct by the federal consumer protection regulator, whose rules cover fair disclosure, data use and recovery practices. Which regime applies depends on your licence, and some lenders sit under both.
The Nigeria Data Protection Act 2023, enforced by the Nigeria Data Protection Commission, governs borrower data. It requires a lawful basis for processing, purpose limitation, data minimisation and appropriate security, and gives borrowers rights over their data, including protection from decisions based solely on automated processing that significantly affect them. A lending app that harvests contacts or photos for collections has a problem under the Act as well as under conduct rules.
AML and counter-terrorist financing duties follow the FATF Recommendations: customer due diligence, ongoing monitoring and suspicious transaction reporting. An agent may surface a suspicious pattern but should not decide whether to report it.
Nigeria has no AI-specific law. Nigerian lenders with UK operations will find the outcome-based approach described in FCA Consumer Duty and AI agents in lending a useful reference for designing fair contact strategies, even where it does not bind them.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
How do you start in the first 90 days?
Month one: start with repayment reconciliation. Unmatched transfers are a daily pain, the right answer is checkable against the bank statement, and the agent's mistakes are reversible before posting. Write down your matching rules first; most teams discover they have three versions.
Month two: shadow the agent on a full month of incoming payments and review every disagreement. Add identity record matching in shadow, with a person clearing every case.
Month three: certify reconciliation at depth 3 within a value limit and keep identity matching at depth 2. Follow the stages in deploying AI agents: train, shadow, certify, widen. For a contrast with an EU lender's rulebook, read our brief on operations agents in French consumer lending.
Where Pryme fits
In Pryme, a salary credit read from the bank's own statement arrives as sealed evidence, while a payslip uploaded by the borrower is declared. The verification agent shows which supported each figure. A screenshot of a transfer sent by a borrower is external and cannot mark a loan as repaid, so the agent holds it until the credit appears in the account.
Questions people ask
Can the agent decide if a mismatched name is fraud?
No. It applies documented tolerance rules and drafts a view, and a person decides. A name mismatch can be a typo, a married name or impersonation, and the consequences of each call are different.
Can an agent talk to borrowers on WhatsApp?
It can answer balance and payoff questions from the loan system, provided the borrower is told plainly that they are talking to an AI. Our piece on FCA Consumer Duty and AI agents in customer service covers the design questions. Collection conversations should go to a person.
What should the agent never have access to?
The borrower's phone contacts, photos and social profiles. Removing that access removes the possibility of debt-shaming by design, and makes your data minimisation case under the Act simple to show.