AI agents for operations leaders in e-commerce in Nigeria

Proving that money really arrived is the first job for AI agents in Nigerian e-commerce operations: matching bank transfers to real credits, reconciling pay-on-delivery cash from riders, chasing failed deliveries and sending vendor agreements for signature. Fraud, rider and vendor disputes and refund exceptions stay with people. The Nigeria Data Protection Act 2023, CBN rules and consumer law apply.

What keeps e-commerce operations leaders in Nigeria busy?

Proving that money arrived. Many customers pay by bank transfer, and some send a screenshot or a forwarded alert as proof. Fake credit alerts are a known fraud, so operations must confirm every transfer against the actual credit in the collection account before releasing goods. With many orders and references that customers type themselves, matching is slow.

Cash in the field. Pay on delivery remains common. Riders and third-party logistics partners collect cash or take transfers at the door and remit later. Reconciling what each rider collected, what they remitted and what is still owed is a daily task, and gaps can become disputes.

Delivery and vendors. Traffic, address descriptions and customers who are unavailable at the agreed time drive failed deliveries. Marketplaces and retailers also onboard vendors and logistics partners, each needing an agreement, a bank account check and ongoing settlement.

Where can AI agents help?

• Transfer confirmation (depth 3). The agent matches each order to a credit in the collection account by amount, sender name and time, marks the order paid only when the credit is in the bank record, and releases it to fulfilment. Screenshots never count as proof. Unmatched orders are held.
• Rider cash reconciliation (depth 2). The agent compares delivered pay-on-delivery orders with each rider's remittances and drafts a daily position by rider and partner. A logistics lead reviews the gaps.
• Failed delivery follow-up (depth 3). The agent contacts the customer from an approved template to rebook delivery, confirms a landmark-based address, and reschedules within the rules. Third attempts are held for a person.
• Vendor agreements (depth 3). Once a vendor passes checks, the agent prepares the standard agreement from the template, sends it for signature and files the signed copy. Our guides on connecting an AI agent to your e-signature platform and letting an AI agent send documents for signature cover the set-up.
• Price change notices (depth 2). Where imported stock is repriced after exchange rate moves, the agent drafts notices to customers with open orders. A person approves before they go out.

Transfer confirmation is the clearest case for a strict source rule. An agent that accepts a screenshot will release goods to fraudsters. An agent that only accepts the credit in the bank's own record cannot be fooled by a fake alert, however convincing.

What stays with people, and why?

Fraud decisions stay human. Blocking a customer or reporting a suspected fake alert to the bank needs a person who has looked at the evidence.

Rider and logistics partner disputes stay with people. A shortfall may be theft, a missed remittance, a customer who paid by transfer to the rider's personal account or a reconciliation error. Each needs a conversation, not an automatic deduction.

Refund exceptions, vendor terminations and agreements with non-standard terms stay with named people.

Which rules apply in Nigeria?

The Nigeria Data Protection Act 2023, enforced by the Nigeria Data Protection Commission, governs customer, rider and vendor data. It requires a lawful basis, purpose limitation, data minimisation, security and respect for data subject rights. Customer phone numbers and addresses shared for delivery should be visible to riders only for the time needed, and the agent should see only what each task needs. Larger processors of personal data may need to register with the Commission and carry out compliance audits.

The Central Bank of Nigeria regulates banks, payment service providers and the payment rails. Disputes over failed or reversed transfers go through the customer's bank and the payment provider under CBN rules, so the agent should route a customer claiming a debit without credit to the right channel rather than promise a refund. Card data must meet PCI DSS v4.0.

Consumer protection law, enforced by the federal consumer protection regulator, requires fair dealing, accurate information and honoured refunds. Price changes on orders already paid for are a particular risk. Nigeria has no AI-specific law; the data protection rules and consumer law are what apply to agents.

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 transfer confirmation. The bank record is the single source of truth, the rule is simple, and fraud losses fall quickly. Agree the matching tolerance for sender names first.

Month two: shadow rider cash reconciliation. The agent drafts each day's position and the logistics team compares it with their own sheet.

Month three: certify transfer confirmation at depth 3 and start sending vendor agreements through the agent. Follow deploying AI agents: train, shadow, certify, widen. For an EU market with very different delivery and refund rules, see our brief on operations agents in French e-commerce.

Where Pryme fits

In Pryme, a credit in the collection account read from the bank arrives as sealed evidence; a customer's screenshot is external. Because an external claim cannot authorise release of goods, the confirmation agent cannot be talked into shipping on a fake alert. Every match and every hold is filed.