Retailers looking to deploy artificial intelligence must first get their data in order. That was the central message from the RETRAK Retail Summit 2026 in Nairobi: before choosing an AI platform, organizations need to organize, label, and link operational data so AI can genuinely improve business decisions.
Why data readiness matters more than AI
Moderated by Moses Kemibaro, a panel including Eric Muriuki (Group Director of Digital Business and CEO of LOOP DFS at NCBA Group) summed up the consensus: it’s “data readiness” rather than “AI readiness.” Retailers already generate continuous streams of information — sales, inventory, suppliers, finance and customer interactions — but these data are often not structured for AI consumption. Without proper preparation, advanced tools risk producing partial or misleading outputs, essentially turning dashboards into chatbots without addressing the underlying fragmentation.
Fragmented sources limit business value
Panelists stressed a pragmatic point: the problem is not a lack of data but its dispersion, which undermines operational visibility. Stores, e‑commerce sites, loyalty programs, mobile apps, social channels, delivery platforms and payment systems often operate in silos. The result: a customer can be known for an in-store purchase with no visible prior online journey, and different departments may work from inconsistent datasets. A unified operational view is therefore a precondition for AI to draw reliable conclusions.
Payments as a source of business intelligence
Judy Waruiru, Regional Managing Director at Network International, highlighted that digital payments are an undervalued repository of insight. Each transaction reveals purchase behavior, preferred channels, timing and customer movement between physical and online stores. Correlated with inventory, loyalty and sales data, payments help spot churn, detect fraud and measure promotion effectiveness. Payment operators, which process transactions across many merchants, can also surface broader consumption trends that a single retailer may not see.
Start with a specific business problem
Panelists advised against overly broad AI programs at the outset. Instead, target a discrete operational use case — replenishment, pricing, fraud detection, campaign recovery or customer retention — and secure measurable gains before scaling AI elsewhere. Siddesh Narkar, Head of Product at Compulynx, recommended focusing on inventory profitability metrics (for example Gross Margin Return on Inventory Investment, GMROI) rather than sales growth alone to determine whether stock is delivering sustainable returns.
Toward customer journeys driven by AI agents
The discussion also looked ahead: Waruiru suggested AI agents could soon make purchases on consumers’ behalf. If that model becomes widespread, retailers will need structured digital storefronts, standardized product information and seamless payment solutions usable by both humans and automated assistants. In that scenario, convenience will compete with price as a primary differentiator.
Concrete actions for the next 12 months
The panel offered practical, convergent steps retailers can take now to prepare for AI:
– Digitize business operations;
– Organize and label existing data;
– Connect customer, inventory, payment and finance systems to build a unified operational view;
– Identify a priority business problem where AI can deliver measurable value;
– Build from concrete results rather than launching a company-wide transformation from day one.
The competitive edge will come less from who adopts AI first and more from who best prepares their data to make AI work. By making information flows coherent and accessible, retailers can turn the promise of AI into measurable operational improvements — and avoid letting AI obscure gaps in data readiness.
