Nigerian collections agents working beside an AI agent must learn to judge what it drafts and holds, to keep recovery actions and contact with third parties in human hands, and to spot the conduct that has drawn regulators' attention to digital lenders. Training is practical, built on the desk's real cases, and takes 90 days.
What fills a collections agent's day at a Nigerian consumer lender?
Most consumer credit in Nigeria is short tenor: salary advances, small personal loans and app-based credit repaid within weeks. Arrears arrive fast and in volume. A collections agent spends the day calling and messaging borrowers whose repayment date has passed, chasing failed direct debits, checking whether a transfer the borrower claims to have made has landed, and agreeing new dates.
The harder part is choosing the next step for each account. The collector decides whether to grant more time, whether to propose a restructure, whether the account should be reported to a credit bureau, and whether to trigger recovery from the borrower's accounts at other banks under the CBN arrangement that allows it. Each of those carries weight for the borrower, and each one is easier to start than to undo.
Where does an AI agent help, and at what depth?
Reading the account before contact (depth 1)
The agent pulls the repayment schedule, failed debit attempts, previous promises and any open complaint into one view. The collector starts every call knowing the facts, and the agent changes nothing.
Drafting reminders and repayment proposals (depth 2)
The agent drafts the reminder or a proposed new schedule within the lender's affordability rules. A person reads and sends it. This is where tone gets checked: a reminder that implies criminal consequences or public exposure must never leave the desk.
Confirming payments and closing promises (depth 3)
When the core banking ledger, a sealed source, shows the payment received, the agent can mark the promise kept and stop further reminders on its own. A borrower's screenshot of a transfer is a declared claim; the agent notes it and holds the account until the ledger confirms.
Bureau reporting and cross-bank recovery (held, depth 4)
Reporting a default or starting recovery from the borrower's other accounts is hard to undo. The agent prepares the case and the evidence, and a named person decides. Many lenders reserve these to supervisors entirely through their own limits.
What must collectors learn, and what must they keep?
The skill gap is reading, not building. Collectors need to know what each verdict means, why an action was refused and which bound decided, and how provenance changes what they can rely on. A borrower who claims to have paid is not lying by default, and a screenshot is not proof by default; the trained collector checks the sealed record. This is the reviewer's slice of AI agent skills by role, and the lender-wide picture is in the competency matrix for the lending team.
Collectors also learn to correct the agent properly. When a draft is wrong, they record why, so the fix lands in the agent's brief and every later borrower benefits.
What stays with the person:
- any contact with someone other than the borrower, including guarantors, unless the contract and the law allow it and a person approves
- any decision to report a default or begin cross-bank recovery
- agreeing a restructure or a waiver
- calls with borrowers in distress or disputing the debt
The longer list of red lines sits in what collections agents should never let an AI agent do.
Which Nigerian rules shape oversight and literacy?
The Central Bank of Nigeria supervises banks, microfinance banks and the other lenders it licenses, and expects them to treat borrowers fairly and to answer for anyone collecting on their behalf. Consumer protection rules aimed at digital lenders go further: they prohibit debt shaming, threats and contacting people in a borrower's phone contacts about the debt. An agent that drafts messages must be bounded so it cannot produce any of these, and the people reviewing it must know these lines by heart.
The Nigeria Data Protection Act 2023, enforced by the Nigeria Data Protection Commission, requires a lawful basis for processing, purpose limitation and respect for data subject rights. For a collections agent this decides which data the AI agent may read, whom it may message and how long records are kept. A borrower's contact list is not collections data, whatever the app was once permitted to read.
There is no dedicated AI law in Nigeria. The obligations fall on the lender as they would for any human or outsourced collector.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
What does a 90-day training plan look like?
- Weeks 1 and 2. Every collector and team lead completes the Foundations material and a one-hour session on the lender's red lines, with real (anonymised) reminders that crossed them.
- Weeks 3 to 6. Shadow review. Collectors decide each held reminder, restructure proposal and bureau referral first, then compare with a team lead's decision. Every disagreement goes into a shared log.
- Weeks 7 to 10. Reviewer training for the collectors who will approve held items, ending in a two-week practicum on live cases.
- Weeks 11 to 13. The lender links approval rights to a current certificate: only certified collectors release held reminders, and bureau reporting stays with supervisors. Weekly calibration begins on the cases where reviewers disagreed most.
Keep which roles in consumer lending need AI agent training in view while planning, since credit and complaints staff touch the same accounts.
Where the Academy fits
Pryme Intelligence Academy starts the whole desk on the free Foundations track and moves reviewers to the Operator track, with tests set in consumer lending. The practicum uses two weeks of real held decisions, needs at least 95 in every 100 correct with no missed red line, and is signed off by a Pryme assessor. Certificates are issued through VerifyDoc.ai and can be checked by number or QR code.
Questions people ask
Can the agent decide to report a borrower to a credit bureau?
It can prepare the report and the evidence, but the decision should be held for a named person. A wrong report is hard to reverse and harms the borrower's access to credit for a long time.
What if a borrower says they paid but the ledger disagrees?
The agent should pause reminders and hold the account for a person, not keep chasing. The collector checks the ledger and the payment channel, then tells the borrower what was found.
Do outsourced recovery agencies need training too?
Yes, if they act on cases the AI agent prepares. The lender answers for their conduct, so they should know the red lines and what a held action means.