The Next Wave of Online Shopping: Where Africa Fits In
Visa says AI shopping agents will become the next phase of commerce, allowing consumers to authorize AI assistants to search for products and complete purchases securely within conversations The company is expanding its tokenization technology to support AI-powered payments, aiming to improve payment security, reduce fraud and increase successful online. . .
Visa says AI shopping agents will become the next phase of commerce, allowing consumers to authorize AI assistants to search for products and complete purchases securely within conversationsThe company is expanding its tokenization technology to support AI-powered payments, aiming to improve payment security, reduce fraud and increase successful online transactionsVisa believes Africa is well positioned to adopt AI-driven commerce, but widespread use will depend on integrating mobile money, strengthening digital security and establishing clear rules on liabilityAfter face-to-face retail and then e-commerce, Visa says commerce is entering a third phase: an “ambient and agentic” era in which artificial intelligence agents make purchases on behalf of consumers. Speaking at the payments network’s annual forum in Paris, Oliver Jenkyn, Visa’s president of global markets, said AI agents would soon shop for both consumers and businesses. Buying Without Leaving the ConversationUnder Visa’s proposed model, users authorize an AI assistant such as ChatGPT with a secure payment credential, along with their preferences and spending limits. The agent then searches for products, compares offers and completes the payment without requiring users to leave the conversation or enter their card details. A traveler, for example, could allow an assistant to book a flight within a predetermined budget, while a small business could have an AI agent automatically reorder supplies. According to Visa, purchases can be completed directly within the conversation. Visa argued that the opportunity lies in a major inefficiency in online commerce. Global digital commerce is worth 4. 8 trillion euros, yet two out of every three online purchases are abandoned before payment, representing 68% of online shopping sessions and about 260 billion euros in lost sales each year, according to figures cited by Mehret Habteab, the company’s head of platform and products. AI is already widely used in the early stages of online shopping. Tareq Muhmood, Visa’s regional president for Africa, the Middle East and Central Europe, said 84% of consumers surveyed by Visa already use AI agents at some point during their shopping journey to search for or compare products, but very few currently allow them to complete the payment. By removing that final point of friction, Visa says AI agents could recover some of those lost sales. Tokenization at the CoreThe system builds on tokenization, a technology Visa has been deploying since 2014 that replaces card numbers with secure, dynamic tokens rather than fixed credentials. According to VisaNet data presented at the forum, between 60% and 70% of Visa’s online transactions are now tokenized. Visa aims to reach full tokenization by 2030, saying the technology has increased successful payment authorizations by 5% and reduced fraud by 35%. Visa is now adding a dedicated layer for AI agents. The system includes next-generation tokens enriched with contextual data, a directory of verified AI agents and merchants, a readiness assessment that allows merchants to determine whether their websites are prepared for AI-driven purchases, and a partnership with OpenAI to secure AI-initiated payments. Each AI transaction carries its own identifier, separate from that of the user, allowing banks to determine whether a payment was initiated by a person or by an AI agent. Three Obstacles RemainJenkyn said three obstacles continue to slow adoption: the technology itself, which is still under development; consumer trust, as users are generally comfortable with AI making recommendations but remain reluctant to let it complete payments; and liability, specifically who is responsible when an AI agent makes an incorrect purchase, whether the user, the AI platform, the merchant or the payment network. He said that final issue, which is both legal and technical, will determine whether the technology achieves mass adoption. On consumer trust, Visa pointed to findings from an internal study presented in Paris. Respondents ranked Visa first among 25 companies when asked which organization they would trust to handle their money in a world of AI agents. Visa also linked that trust to cybersecurity. The company said in Paris that it is one of about 40 critical infrastructure companies given access by Anthropic, the developer of the Claude AI assistant, to its advanced Mythos model, whose vulnerability detection capabilities are considered too sensitive for public release. Visa said it tested the model against its own security systems over several months and found that its defenses successfully withstood the testing before open-sourcing part of its vulnerability remediation toolkit. Can Africa Keep Pace as the Global Leader in Mobile Money?Asked whether Africa risked being left behind by a technology initially designed for markets with widespread card use and e-commerce, Aminata Kane, Visa’s head of West and Central Africa, said the continent was already better positioned than many assume. She said most African fintech companies already use artificial intelligence to prevent fraud or provide customer support through chatbots. According to Kane, the next step is expanding tokenization across the payments ecosystem. Visa said it has already begun working with mobile money operators and banks across the subregion and has launched tokenization with banks in Ghana. She added that Africa would not miss the opportunity. Africa’s path, however, could differ from that of more mature markets. The agentic commerce model presented in Paris is built around payment cards, whereas mobile wallets remain Africa’s primary payment method. According to the GSMA, the continent accounts for two-thirds of the global value of mobile money transactions, representing about $1. 4 trillion exchanged in sub-Saharan Africa in 2025. Whether AI agents will eventually be able to pay merchants in Cotonou or Nairobi as easily as those on U. S. e-commerce sites will depend on the ability to link tokens and virtual cards to mobile money accounts. Visa already does this with Wave, whose accounts can be linked to a virtual card, with Orange Money through a partnership established at the end of 2025, and through card-issuing fintechs such as Côte d’Ivoire’s Djamo. Another factor is that conversational commerce already exists across much of Africa in a more informal form through sales conducted on WhatsApp, TikTok and other social media platforms. Agentic commerce may, therefore, formalize an existing shopping habit rather than create an entirely new one. Commercial rollouts have already begun in other markets across the region, and banks are preparing for them, but no specific timeline for Africa was announced in Paris. Between ongoing tokenization efforts in Ghana, growing interest from fintech companies and the cautious approach of regulators, the pace of AI-driven commerce in Africa will likely depend on three areas: securing digital credentials, integrating mobile money and clarifying liability. The third phase of commerce may already be taking shape, but its African version has yet to be written. Fiacre E. Kakpo