Contents
Most payment news is addressed to consumers: a faster app, a cheaper transfer, a new QR code at the till. This one is addressed to software. The central bank of the eight WAEMU countries has published the first official list of programming interfaces that let a company’s own accounting, invoicing or treasury system put payments onto the regional instant payment rail and read them back — and it arrives with a connection deadline that falls this month.
What is happening?
On 16 September 2026 the Banque Centrale des États de l’Afrique de l’Ouest (BCEAO) issued a communiqué announcing the homologation — formal approval — of Business APIs for PI-SPI, the interoperable instant payment platform of the West African Economic and Monetary Union. The annexed list, updated 17 September 2026, names 24 approved interfaces across all eight member states: Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal and Togo.
The distribution is uneven. Senegal accounts for nine of the 24 and Côte d’Ivoire for eight; the remaining seven are spread one or two each across the other six countries. The providers are a mix of banks — including Ecobank, which appears in most of the country lists, and Standard Chartered — the mobile financial services arms of Orange and MTN, and specialist payment firms. The BCEAO says the list will be updated as further participants complete homologation.
What an approved Business API does, in the central bank’s own description, is let a business issue and receive payments, submit bulk settlement requests, automate secured payment operations, follow transactions in real time and improve its cash management. In plain terms: the payment is initiated by the company’s software rather than by a person on a banking portal, and the confirmation returns to that same software.
PI-SPI itself launched on 30 September 2025 in Dakar. It runs continuously across the eight countries and settles in seconds. Connection is not optional for the institutions: a BCEAO communiqué of 2 April 2026 set a deadline of 30 June 2026, and a further communiqué of 25 June 2026 extended it to 30 September 2026 for banks, electronic money institutions and payment institutions, with microfinance institutions supervised by the Banking Commission given until 30 June 2027. At 24 June 2026, 80 participants were connected and 74 more were in live testing.
Two things remain uncertain, and both matter. The BCEAO has not published PI-SPI transaction volumes or values, so actual usage is not in the public record. Nor has a fee schedule for business or API-initiated payments been published: the launch material describes free, instant access to funds for individuals and standardised QR acceptance for merchants, but what a company pays to collect or disburse through an API is a commercial matter with its provider.
Why it matters economically
The economic weight of this is not in the transaction fee. It is in time and in labour.
Consider where value actually moves in the union. According to figures from the BCEAO’s 2025 payment systems reporting, cheques represented 24.6% of the volume passing through the regional clearing system that year but 71.7% of its value, while electronic transfers made up 74.1% of volume. The small instruments have gone electronic; the large ones have not. The payments that decide whether a business can pay its own suppliers this week are still the slowest in the system.
Every day between “the customer has paid” and “the money is usable” is working capital the business has financed itself. Take an illustrative distributor in Abidjan with CFA 40 million of monthly receivables settled by cheque and a three-day average clearing cycle: roughly CFA 4 million sits in transit at any moment — money that cannot buy stock, cover payroll or reduce an overdraft. Instant settlement does not add a franc of revenue; it releases capital the business already earned. Our view is that this, not any fee saving, is where the arithmetic is decided.
The second effect is on the cost of knowing. A payment initiated by your own system carries your own reference. It returns to the system that raised the invoice. The match is made when the payment is created, not reconstructed afterwards from a statement. That removes a category of work rather than speeding it up — which is a different economics from buying a faster tool.
What it means for African businesses
This applies to the eight WAEMU states and to nothing else. It is not a continental development: Nigeria, Ghana and every market outside the union have their own rails on their own timetables, and a business trading across those borders still faces the currency and settlement questions set out on our markets page.
Within the union, three groups are affected differently.
Businesses with large B2B receivables — distribution, logistics and trade, construction supply, professional services — stand to gain most, because they are the ones still taking cheques and waiting. The logistics and trade pattern is the clearest: goods release is gated on payment confirmation, and confirmation is the slow step.
Businesses with high-volume consumer collections gain less from the API than from PI-SPI itself, and many already run on mobile money. For them the problem is the one we examined in mobile money is a payment method — reconciliation is the business system, and a Business API attacks it at the source rather than at the statement.
Businesses that pay out in bulk — supplier runs, agent commissions, per-trip driver settlements, refunds — get something they have not had: a programmable disbursement channel inside the union, with confirmations that return to the system that authorised them.
The binding constraint is not regulation. It is which institution you bank with. Approval of an API does not oblige a provider to offer it to you, on terms you can accept, with documentation you can build against. And 30 September is a deadline for institutions to connect to PI-SPI — not a promise that yours will expose a Business API to its corporate customers.
Technology and AI implications
It is worth being exact about what technology does and does not solve here.
An API is useless without a system of record at the other end. If invoices live in a spreadsheet and customer history lives in a messaging thread, there is nothing for the payment confirmation to attach itself to. The integration is the last step of building a business system, not a substitute for one.
AI is largely irrelevant to this problem, and it is worth saying so plainly. A payment that carries its own invoice reference needs no intelligence to be matched; it needs a field. Using a language model to guess which invoice a payment belongs to, when the rail can carry the answer, is a worse system dressed as a better one. AI earns its place only on the residue: payments that arrive without a reference, partial and combined settlements, customers claiming a payment that has not landed, and deciding which overdue accounts to chase. That residue is real, but it is the minority of the work, and it shrinks as the rail is used properly.
What implementation requires is unglamorous: an accounting or ERP system that can hold an external payment identifier; a provider that will contract with you for API access; key management handled by someone accountable; dual authorisation on anything that moves money out; and a tested path for what happens when a call fails or a confirmation is delayed. The BCEAO operates a public sandbox for the Business API, so the specification can be examined before any commercial commitment is made.
Practical business implications
Four questions are worth putting to your bank or payment provider this month.
- Are you connected to PI-SPI, and do you hold an approved Business API? Both, not one. Check the answer against the BCEAO’s published list rather than a sales conversation.
- What does it cost us to collect and to disburse through it? Ask for the schedule in writing, including per-transaction and per-batch charges, and note that no public tariff exists for business use.
- What does the confirmation contain? Specifically: can it carry our invoice or order reference, and is it returned in real time or in a file?
- What are the limits? Per transaction, per day, per batch — these decide whether the rail is usable for your largest payments or only your smallest.
Then two for yourself. Does the software that raises your invoices have somewhere to put a payment reference and a settlement status — if not, that is the project, not the API. And which single process would you automate first? Our view is that bulk disbursement is the better starting point: the volumes are predictable, the counterparties are known, and the failure modes are visible immediately. If your provider offers nothing, that is information too — a reason to review the banking relationship, not to build a workaround.
RevenueStack perspective
Payment rails commoditise. That is their purpose: a central bank builds one so that no single institution can charge rent for moving money between others. Within a few years, instant settlement across the union will be unremarkable, and no business will win a customer by having it.
The advantage does not sit in the rail. It sits in the system that can speak to it. When a payment arrives already attached to an order, the business gets what the rail alone cannot give: goods released the same hour, a credit decision made on a live balance rather than last month’s statement, a finance team whose day is not spent matching. That is the Strategic Economics point — the infrastructure is public and shared, so the return accrues to whoever has built the private system on top of it. Firms with a real operating system will convert this into cash cycle and margin; firms without one will experience it as a slightly faster bank portal.
There is a quieter implication. Twenty-four approved interfaces across eight countries is thin supply — not a failure, an early market. It also means the constraint has moved. For a decade the answer to “why can’t our software just pay this supplier?” was regulatory and infrastructural. In the WAEMU countries, from this month, it is increasingly a software question — one a business can act on, through the integration work described under AI and automation and the record-keeping systems under CRM and business systems.
Conclusion
A central bank has made it possible for ordinary business software to move money instantly across eight countries, and has published the list of institutions cleared to offer it. The businesses that benefit will not be the ones that adopt the rail fastest; they will be the ones that already know, in a system rather than in someone’s head, what every payment is for. For everyone else the useful step this month is smaller: find out whether your provider is on the list, ask what it would cost, and look hard at whether your own records could receive the answer.
Sources
- Homologation des API Business de la Plateforme Interopérable du Système de Paiement Instantané (PI-SPI) de l'Union Economique et Monétaire Ouest Africaine — BCEAO, 2026-09-16
- Liste des API-Business PI-SPI homologuées par la BCEAO (mise à jour 17 septembre 2026) — BCEAO, 2026-09-17
- Prolongation du délai de connexion à la Plateforme Interopérable du Système de Paiement Instantané (PI-SPI) de l'Union Economique et Monétaire Ouest Africaine — BCEAO, 2026-06-25
- Connexion à la Plateforme Interopérable du Système de Paiement Instantané (PI-SPI) de l'Union Economique et Monétaire Ouest Africaine — BCEAO, 2026-04-02
- Lancement officiel de la Plateforme Interopérable du Système de Paiement Instantané (PI-SPI) de l'UEMOA — BCEAO, 2025-09-30
- PI-SPI Sandbox — Environnement de test de l'API Business — BCEAO, 2026-09
- 71% of WAEMU Transaction Value Still Flows Through Cheques — Despite BCEAO's New Instant Payment Rail — Launch Base Africa, 2026-07-24
- UEMOA : 24 solutions permettent désormais aux entreprises d'utiliser le paiement instantané — Agence Ecofin, 2026-09-17