Paystack has launched an experimental product that lets AI chatbots pay for things on user behalf — the latest sign that Africa’s fintech giants are also racing to figure out what happens when agents, not apps, become the checkout counter.
The new experimental product is called Paystack Index, and it went live this week in Nigeria. Instead of opening an app to top up your phone, send money, or order dinner, the new service allows you to ask an AI agent to do it instead. The agent communicates this intent to Index, Index links to Paystack’s payment rails, and the transaction completes inside the chat window.
What Paystack Index can do
Paystack Index is built by TSG Labs, the experimental venture studio that sits inside The Stack Group — the holding company Paystack created in January 2026 to house its core payments business, its consumer wallet app Zap, and Paystack Microfinance Bank. Index is TSG Labs’ first product to reach the public.
Here’s the mechanic: a user tells a supported AI agent to do something, e.g. “top up my MTN line,” “send ₦5,000 to my sister,” “order jollof rice from Chowdeck,” and Index interprets the request, routes it to the right merchant, and processes the payment through Paystack and Zap’s infrastructure. The checkout happens without the user ever leaving the chat.
At launch, three AI clients are supported: ChatGPT, Claude, and OpenClaw (the open-source personal AI agent that runs on messaging apps like WhatsApp and Telegram). Four use cases are now live:
- Buying airtime and mobile data
- Funding a wallet
- Sending money through Zap
- Ordering food via Chowdeck
Paystack is keeping the merchant list short on purpose. CEO Shola Akinlade has framed Index as an experiment, not a full-scale rollout yet. He’s explicitly told Techpoint Africa he doesn’t expect meaningful usage for at least another year and the company will now just learn about early adopters’ preferences and agent use patterns.
“The goal is to learn how people want to use AI agents for everyday commerce, how merchants can show up in these new experiences, and what infrastructure is needed to make AI-led checkout work safely for African markets from the beginning,” Akinlade said.
Users stay in control through spending limits and permissions they configure after signup, and Paystack is the one deciding which merchants get plugged in for now. Though Paystack supports 300,000 businesses across Africa, the Index participants list remains limited and carefully curated. So do the use cases, most plain and routine ones, the food ordering remaining, perhaps, the biggest challenge for AI agents use.
The bigger bet behind it, in Akinlade’s words, is that agent search is soon to become a default one rather than secondary alternative: “Agents are coming up and browsers are going down.” Paystack wants the agentic payments rails ready before that shift actually happens at scale.
So how big is the Index target audience right now?
The honest answer is that the current potential user base for Paystack’s new product is relatively small. However, it doesn’t mean that its future adoption levels will also be minimal.
At present, the Index is available in Nigeria only. Nigeria has 182.2 million active telecom subscriptions and roughly 151.6 million people connected to the internet, the vast majority through mobile data. That’s the absolute outer limit of who could ever discover an AI-checkout product in the country. At the same time, if proven successful, Paystack’s innovative product should also expand to Ghana, Kenya and South Africa.
If we drop down to people who already trust a digital wallet with their money, the number shrinks fast, though it’s still large. OPay alone serves an estimated 40 to 50 million registered users in Nigeria. PalmPay has around 35 million. These are the people who are potentially comfortable with handing money over to an app, though not yet an agentic one.
Paystack has never published a user count for Zap, its own consumer wallet — the product Index leans on most heavily for wallet funding and transfers. Although Zap runs on infrastructure that already serves over 300,000 businesses across Africa, its own actual user data over the first year of operations is currently not known.
Chowdeck ordering is at present the most complex use case. Food delivery is the only multi-step transaction Index currently supports, and here we have actual numbers to estimate the potential user base. The service crossed 2 million registered users across Nigeria and Ghana by December 2025, doubling from 1 million in under a year. In Nigeria alone, the user base is likely somewhere around 1.5 to 1.8 million people who may opt for the same service via popular chatbots as well.
The final question is: who’s using these AI agents in Nigeria? This one is tricky, as most AI chatbot providers do not publish such data. At the same time, we can make some rough estimations.
- DataReportal’s mid-2025 Global Statshot put ChatGPT adoption among Nigerian internet users aged 16+ at just 8.2% — well behind Kenya’s 42.1% and South Africa’s 15.3%. Applied loosely to Nigeria’s internet population, that’could equal to about 12 million people who’ve used ChatGPT so far.
- Claude has no published Nigeria user count, but in Anthropic’s own Economic Index, Nigeria’s “AI Usage Index” sits at just 0.2 — meaning Claude usage is one-fifth of what the country’s working-age population share would predict. That puts Nigeria behind India (0.27) and Indonesia (0.36) on the same scale. Claude’s Nigerian base is almost certainly a small fraction of ChatGPT’s.
- OpenClaw is a self-hosted tool with no official download numbers, an estimated few hundred thousand global enthusiasts. Its Nigerian footprint is realistically in the thousands, not millions.
Once we stack all four filters on top of each other: internet access, wallet trust, merchant familiarity, and AI agent usage, it becomes clear that the people who could plausibly use Paystack Index today land in the low single-digit millions. That’s a decent user pool to start the testing, though it may seem a small share of the total local population.
This isn’t the first AI-payments move in Africa, but the first consumer-facing one
Paystack Index is genuinely new in one specific way: it’s the first product built for ordinary Nigerian consumers to transact through an AI agent, rather than for developers to build AI features into their own apps. Though several Paystack’s African competitors already test agentic capabilities, the rest of the continent’s “AI meets payments” activity so far has been squarely developer-facing. Here are some examples.
Anchor, a Lagos-based banking infrastructure provider, launched what it called Nigeria’s first fintech MCP server earlier this month. It lets AI coding tools like Claude and Cursor pull live, accurate API documentation when developers are building payment integrations — solving a real problem (AI tools hallucinating outdated API specs) but aimed entirely at engineers, not shoppers.
Flutterwave has a community-built MCP server too, that exposes its payment API as tools an AI agent can call, i.e. create a payment link, verify a transaction, issue a refund. Again, this is a developer’s tool for wiring AI into business workflows, not a consumer checkout experience.
Yoco, the South African payments company, rolled out its own MCP server so merchants can ask an AI assistant to generate a payment link or reconcile accounts by typing a sentence instead of opening a dashboard. Same pattern: built for the business owner, not the end customer.
Afriex, a cross-border remittance startup, built an MCP server that lets an AI agent chain together African currency conversions and payouts — useful for someone automating payroll across multiple countries, not for someone asking an agent to order lunch.
There’s even a community-built “Africa Payments MCP” that bundles M-Pesa, Paystack, and MTN MoMo into one interface so a developer can prompt an AI coding assistant in plain English instead of juggling three different SDKs.
What none of these do (and what Paystack Index does) is let a regular person’s AI agent actually complete a purchase on their behalf, end to end, with money moving and a receipt landing in a chat thread. Globally, China’s Alipay is the closest comparison: its AI Pay product reportedly hit 100 million users by February 2026, complete with its own payment MCP server and pre-built AI-tipping and subscription flows. Paystack is now testing whether a smaller, much earlier version of that idea can work in Nigeria.
“The thing we know is that there’s a bunch of early adopters in Nigeria or in Africa. There are people that used ChatGPT on the first day and back to the spirit of TSG Labs, we wanted to start building with those early adopters,” said Akinlade.
The audience size estimate above isn’t really a verdict on whether Index will get a fair share of customers. It’s a snapshot of how early this category still is everywhere, not just in Nigeria. Akinlade’s own comparison — agentic commerce as a shift on the scale of brick-and-mortar retail moving online, is the right frame for judging this, not week-one transaction volume.
