E-commerce companies may struggle to get meaningful returns from AI unless they first fix their data infrastructure. Product information, inventory records and customer data are often scattered across different systems, preventing AI from developing a complete picture of the business. The report identifies data readiness as one of the biggest factors determining whether AI projects move beyond experimentation and deliver measurable business outcomes.
The problem becomes particularly serious when AI is used for personalisation, pricing, forecasting and inventory management. Incorrect inventory counts or duplicated customer records can be amplified by AI, potentially producing unreliable recommendations and forecasts at much greater scale. For example, an AI marketing system that is not connected to accurate inventory data could recommend or promote products that are unavailable.
Nisum therefore recommends unifying fragmented data sources, establishing clear ownership of data quality and introducing governance before scaling AI initiatives. It also argues that data preparation cannot be treated as a one-time exercise: products, suppliers, inventory and customer behaviour constantly change, meaning AI systems need continuously updated and consistent information.
The broader takeaway is that the biggest obstacle to AI commerce may not be the AI model—it may be the data underneath it. Retailers can invest in sophisticated generative AI and automation, but those systems cannot compensate for fragmented or unreliable business information. As e-commerce moves toward more autonomous AI-driven operations, the competitive advantage could increasingly belong to companies with clean, centralised and well-governed data, not simply those using the most advanced models.
