Nigeria markets chase AI spend as readiness lags
KPMG asked 176 African chief executives where their money is going. The answer: artificial intelligence. The same survey, part of a global poll of 1,400 CEOs, also found a readiness gap. Read that pairing carefully. In Nigeria markets, companies are funding AI ambitions with budgets built for a different operating reality.
The gap is not enthusiasm. It is the unglamorous inputs: clean customer records, reconciled ledgers, and staff who can tell a bad model output from a good one.
Identity data is the weak link
Nigerian lenders and fintechs want AI for credit scoring and fraud detection. Both depend on KYC files. The CBN's tiered KYC rules let customers open accounts with minimal identification. That design widened financial access. It also leaves holes in training data where identity should be.
A model trained on thin identity data learns the gaps. Fraud rings learn them faster. Per the KPMG survey, CEOs see the readiness shortfall and are spending anyway. This suggests AI fraud tools will reach the institutions with the messiest data first, which is the reverse of where they would help most.
Then there is the NDPC. Training a model on customer data needs a lawful basis under the Nigeria Data Protection Act, plus documented records of it. Most AI pilots start in a sandbox run by the innovation team, not the compliance team. Expect enforcement questions after deployment, not before.
Agent float is the live test case
Agent banking is where AI promises meet cash. Forecasting agent liquidity needs is a textbook use case: transaction history in, cash demand out.
The catch sits under the forecast. A large share of agent transactions still get recorded on paper or reconciled days late. Feed a model a partial transaction picture and it will under-order cash. Trim float buffers on that forecast and agents run dry on the days that matter: market days, salary weeks, school fee deadlines.
The risk is not a bad forecast. It is a trust collapse inside an agent network that took years to build. Agents who cannot cover a day's withdrawals move to a competitor or stop trading. AI does not fix that. Reconciliation does.
Who pays for the gap
Vendors of cloud capacity and data tooling collect first. Systems integrators collect second. CEOs who fund pilots from innovation budgets while the core ledger still closes manually collect nothing.
The 176 responses describe intent, not execution. Intent is cheap. This suggests the next two years will produce a long list of stalled pilots and a short list of institutions that fixed their data plumbing before buying models.
Here is the blunt version. Nigeria markets do not have an AI readiness problem. They have a records problem, and AI makes the cost of it visible.
Expect boards to start asking a question they have dodged: which of our systems can produce a complete, auditable transaction record for the last ninety days? Where the answer is none, the AI budget is a bet on someone else's infrastructure.