When AI Excludes the Market it Seeks to Modernise

The real risk in African retail is not AI adoption. It is using AI that does not take into account how people actually buy, negotiate, trust and deal.

The conventional wisdom is that improved technology leads to better markets. In African retail, this is only half the story. If AI solutions do not account for how people trade, build trust, manage data costs, and buy through social discussion, then technical development would face delayed adoption.

Take Akin, a vendor of clothes in Balogun market, Lagos, for instance. His business does not revolve solely around stock availability and pricing. It is a matter of discussion, intimacy, seduction and bargaining. In his world, haggling is not a market inefficiency. It is part of how buyers evaluate trust, exercise agency, and decide what a product is worth. Many of the AI technologies coming into African retail are being created as though these habits should be taken away rather than understood.

Now, that is the real blind spot. Companies across the region are investing in AI to improve consumer interaction, forecasting, retail and inventory management. The potential is real as well as the mismatch. Too many AI implementations are still based on imported assumptions about user behaviour, infrastructure, and what constitutes a valid transaction. These assumptions are untrue for much of African retail.

Inclusive AI for Africa

If AI solutions are intended only for metropolitan clients with reliable internet access, official records, and consistent digital behaviour, they will overlook a large segment of the market that still relies on cash, debate, and flexible exchange.

“Design determines inclusion.

So, new technology is not the only thing to be dealt with. It fits, technically. Informal retail systems in Africa are shaped by bargaining, multilingual communication, limited data constraints, relational trust and social proof. If AI doesn’t consider these features, it doesn’t modernise the market. It is wrong.

Why bargaining is important

A significant example is price bargaining. Many African customers still expect some kind of negotiation and often feel they have more power over price in open markets than in digital platforms. That’s key because flat-price AI solutions may appear economical from the firm’s perspective, but seem rigid and alienating from the shopper’s perspective.

In these situations the negotiation is not the commotion around the sale. It’s part of the logic of the sale. Bargaining signals buyer seriousness, compares value, and builds social interaction with sellers. If corporations want AI to work in these marketplaces, the goal should not be to get rid of bargaining. It should be about structuring it in ways that protect customer agency and improve business decision making.

That means a more productive route to innovation. Retailers can construct systems that support dynamic packages, guided negotiation and culturally aware counteroffers rather to just building bots that push fixed offers. India’s Hagglebot, built using Google Assistant and Flipkart, is a useful proof point that AI-mediated negotiating can be made technically possible and financially engaging in a haggling culture.

“Many African customers still expect some kind of negotiation and often feel they have more power over price in open markets than in digital platforms.

Design for the hustle

The second gap is the reality of infrastructure. Even where network coverage has expanded, many African users are still hampered by device access, data costs and sporadic connectivity. So the most technically stunning AI system might not be the most commercially useful system.

AI design in these sectors begins with little friction, not the greatest sophistication. Retail tools should be lightweight, mobile first, multilingual and resilient in low data conditions. And they need to be about the bite-size logic of spending and selling that currently affects informal commerce. The analogy of the sachet economy is applicable here. If consumers want smaller and more flexible units, AI services may likewise need to be modular, inexpensive and easily accessible in portions rather than as a full stack.

Here is where numerous potential instruments are heading. Lengo AI has been defined as mapping informal retailers in West Africa with AI and field intelligence with a friction-light methodology that incorporates WhatsApp onboarding and geotagged photographs to bring small businesses into the view of brands and distributors. This is important because it starts from the point that many informal outlets are either under-recorded, poorly digitised or effectively invisible in official statistics.

“Retail tools should be lightweight, mobile first, multilingual and resilient in low data conditions.

VOA reports a similar lesson from a retail chatbot located in Ghana. The system enables ordering, mobile money payments and streamlined SMS updates and is being extended with voice recognition for non-literate users. That combination better fits the needs of low-literacy, mobile-first and payment-fragmented markets than the traditional app-heavy models brought from more formal retail environments.

AI must be trusted

The third gap is trust, and it may be the most essential. Shoppers might trust AI-generated information, but they may wonder if AI knows what is fair, reliable or safe in a transaction. That’s an important difference. Trust in information is not trust in transaction.

Merchants can not stop at product recommendations or customer support. It has to help make trust visible. Verified seller tags, detailed justifications, community ratings, complaint mechanisms and clear escalation procedures are not extras. They are crucial to determining whether digital commerce is safe enough to implement at all in many African retail contexts.

This alters the role for retail AI. And it should also work as a trust accreditor and not just as an assistant. That is not to say that we should replace human judgment. It involves deploying digital systems to strengthen the signals of trustworthiness, accountability and recourse that buyers already seek when they purchase.

This is a drawback because many mainstream technologies are built for organised retail environments with clearer data, stronger formal records and more uniform indications of trust. Commentary on AI in retail has warned that recommendation, profiling and pricing algorithms can perpetuate prejudice when trained on narrow or unrepresentative data. In practice, that means a system can look intelligent and still fail to grasp the context of low-income, informal or trust-sensitive clients.

“Street-smart AI involves deploying models to strengthen the signals of trustworthiness, accountability and recourse that buyers already seek when they purchase.

Commerce as Social

A large part of trade in African marketplaces is now taking place via conversational channels, mainly WhatsApp and other social networks. Referrals, tales, group discussion and informal endorsement are often the basis for purchase decisions. Shopping is hardly a solo sport.

This is why conversational AI makes such a difference. The future is less about retail tools that live outside the normal flow of communication and more about assistants that live inside the normal flow of conversation. Jumia said some of its most popular AI applications are virtual assistants and conversational tools on WhatsApp and social media, adding that African e-commerce also requires mobile-first, low-bandwidth, language-aware and trust-sensitive design.

“The future will be shaped by assistants that live inside the normal flow of conversation, not retail tools that live outside the normal flow of communication.

This social aspect also accounts for the inability of purely optimisation-led systems to dominate the market. Recent tests of shopping-assistant base models showed that some systems tend to drift towards premium-biased product recommendations unless stringent financial constraints are enforced. That’s a helpful caution for African retail, where many buyers are highly price sensitive and where AI that nudges upward on price can silently undermine trust instead of building it.

What Smart Companies Do

African retailers should stop asking how AI can erase the roughness of informal marketplaces and start asking how AI can analyse the intelligence within them. Informal retail typically appears chaotic from the outside, but much of this seeming disarray is patterned human conduct with commercial meaning.

The companies that will benefit most from AI are those who construct what may be dubbed street-smart systems. These are systems that promote bargaining, rather than erase it; adapt to constrained infrastructure, rather than presume abundance; make trust visible, rather than implicit; and strengthen social commerce, rather than flatten it.

African retail may therefore have a message well beyond the continent. As AI starts to operate in more diverse, unequal marketplaces, success will be less about the size of the model and more about whether the technology can operate intelligently in human systems that are negotiated, social, multilingual and trust-sensitive.

The firms that prevail in that future won’t only have more sophisticated AI. “They will have AI that has learned how the market really talks.

This article draws on insights shared in Uchenna Uzo, Ephraim Nwopkoro and Sheneni Maiyaki’s chapter titled “The AI Puzzle” in the Africa Retail Academy book Scaling Smart: AI, Retail and the Imminent Disruptions.