When Walmart — a company with more engineering resources than most African nations have GDP — publicly admits that its AI checkout experiment converted at one-third the rate of its own website, founders across Lagos, Nairobi, and Accra should be paying close attention. Not because the technology failed. But because of why it failed.
The lesson isn’t about ChatGPT. It’s about trust, infrastructure, and who actually controls the customer relationship.
What Happened, Briefly
Walmart made around 200,000 products available through OpenAI’s Instant Checkout, allowing users to complete purchases inside ChatGPT without ever leaving the chat. It sounded like frictionless commerce. Daniel Danker, Walmart’s EVP of Product and Design, said conversions came in at just one-third the rate of click-out transactions — where users were redirected to Walmart’s own website. He called the experience “unsatisfying.” OpenAI is now scrapping Instant Checkout entirely.
Walmart is replacing it with Sparky, its own chatbot, embedded inside ChatGPT — routing users back to Walmart.com or its mobile app to complete transactions, rather than finishing inside the chat.
The full circle is the lesson.
1. Friction Isn’t Always the Enemy
The dominant narrative in African fintech for the last decade has been: remove every possible step between intent and purchase. One-click. USSD. Tap-to-pay. The assumption is that friction kills conversion.
Walmart’s data challenges that assumption at the margins. The core problem with Instant Checkout was that it forced single-item purchases — every product triggered its own individual transaction, its own shipping calculation, its own delivery. Consumers instinctively understood that five product recommendations meant five separate deliveries. They opted out.
For African founders building commerce or payments infrastructure, this is a reminder that the right friction — a clear cart, a recognizable checkout page, a familiar brand environment — can actually reassure a customer rather than lose them. Especially in markets where digital trust is still being built. A Kenyan shopper who has used M-Pesa for eight years converts better on M-Pesa than on a slicker interface they’ve never seen before. Don’t strip out the familiar in pursuit of the futuristic.
2. Discovery and Conversion Are Different Jobs — Build Accordingly
ChatGPT is now bringing Walmart approximately twice the rate of new customers as search engines. And yet, Walmart is actively routing those customers away from ChatGPT to complete their purchases. That’s not a contradiction — it’s a strategy.
AI platforms are proving to be extraordinary discovery environments. They surface the right product, answer the right question, and reduce the research burden for a buyer. But they are not — yet — trusted completion environments.
African founders should map their funnels accordingly. If you are building a marketplace, a logistics platform, or a consumer app, think carefully about where AI fits. It may belong at the top — as recommendation engine, as customer service layer, as onboarding guide — while your own interface handles the money. Conflating the two jobs is what Walmart did, and what cost them conversions.
3. Owning the Customer Relationship Is Non-Negotiable
This is the most important lesson, and it’s one African startups learn the hard way when they build too deep on top of someone else’s platform.
Rather than a checkout process embedded in ChatGPT’s interface using OpenAI’s payment infrastructure, Walmart’s new model has consumers interact with Sparky — a Walmart-controlled environment running inside another company’s chat interface, with baskets from Walmart’s website, the Walmart app, and ChatGPT all synchronized.
Walmart got its products inside ChatGPT, but kept the checkout, the data, the relationship, and the repeat purchase on its own rails. That’s the architecture that matters.
For founders on the continent: whether you’re building on WhatsApp Business, Instagram Shopping, or any third-party AI interface, the customer data, the transaction record, and the re-engagement capability must flow back to infrastructure you control. The moment you cede the checkout — and with it, the purchase history, the behavioral data, the loyalty loop — you’ve handed your most valuable asset to someone else’s platform.
Flutterwave didn’t become a billion-dollar company by letting Facebook own its payment flows. Paystack didn’t win by being fully embedded in someone else’s checkout. Own the rails.
4. Average Order Value Tells You More Than Conversion Rate
Walmart US CEO David Guggina disclosed that Sparky users have an average order value approximately 35 percent higher than non-Sparky customers. The AI wasn’t converting more people — it was converting better people, spending more per session.
This is a critical metric for African founders to watch as AI tooling becomes more accessible across the continent. In markets where disposable income varies dramatically across customer segments, an AI interface that filters toward higher-intent, higher-value customers is worth building for — even if raw conversion numbers look modest. Don’t optimize solely for volume. Know what your AI channel’s customer looks like versus your organic channel’s customer. The profiles may be radically different, and your monetization strategy should reflect that.
5. The Infrastructure Gap Is Actually a Positioning Opportunity
The reason Walmart’s own surfaces outconvert ChatGPT’s checkout is largely trust and familiarity — consumers who already have Walmart accounts, saved addresses, and payment methods complete purchases in Walmart’s own environment.
Most African markets don’t have that depth of consumer digital infrastructure yet. Fewer saved cards. Fewer stored addresses. Lower digital purchase history. That sounds like a disadvantage. It isn’t — not entirely.
It means that for African startups, the race to become the trusted surface is still wide open. The company that builds the checkout layer that consumers in Lagos or Kampala trust the way Americans trust Amazon is still being built. The wallet that becomes the default. The marketplace that feels safe. That company will win the same battle Walmart is fighting — except it won’t have to fight back from someone else’s platform to reclaim it.
The lesson from Walmart isn’t that AI commerce doesn’t work. It’s that trust compounds, platforms are rented land, and the brand that owns the customer’s last tap wins. For founders building on the continent right now, that last tap is still up for grabs.
