A recent NielsenIQ study says that 51% of US consumers now use artificial intelligence (AI) to shop. Earlier this year, Deloitte released a study of 13,500 European consumers, which found that more than half already use AI for shopping, another 27% intend to, with Spain, Poland, Hungary and Italy above 60% and the UK, France and Germany at 48–52%. There are many reports and studies that say the same thing: globally, retailers are scrambling to keep up with consumer demand for AI-assisted shopping.
With this in mind, as South African retailers race to deploy automated shopping assistants, the primary enterprise challenge is no longer who can launch an AI interface the fastest, but whether the data, architecture, and governance beneath it can be trusted, says Dewald van Staden, redPanda Software’s Head of Engineering. redPanda Software is a trusted technology partner dedicated to developing highly customised software solutions for leading retailers.
He says that rushing to automate the shopping basket without rock-solid data quality and human accountability risks building shiny digital features on weak operational foundations. “The retail shopping experience is becoming increasingly invisible, which in turn makes visible human oversight non-negotiable,” he says.
He argues that as AI moves from helping shoppers make decisions towards making decisions on their behalf, retailers must pay closer attention to transparency, consent and accountability. “AI shopping assistants may make shopping more convenient, but a quick recommendation is not necessarily accurate, relevant or appropriate for the customer. This is because an invisible shopper or AI interface is only as intelligent as the data, integration and governance behind it,” says Van Staden. “Retailers should not simply begin with a shiny AI interface and hop on the bandwagon. Speed and rapid responses do not prove that a recommendation is accurate, relevant, or appropriate – and this builds trust.”
Van Staden outlines three foundational pillars that retailers must establish before deploying AI-driven shopping tools:
- Trusted Data: AI relies entirely on the quality of the underlying data. Inventory records, pricing, and product details must accurately reflect real-world stock, he says.
- Responsible Governance: He says that retailers have to maintain absolute clarity regarding what customer data is used, for what specific purpose, with what consent, and under whose accountability.
- Operational Readiness: “The customer-facing AI layer must be backed by resilient back-end systems,” he explains. Simply put, supply chains, promotions, order fulfilment, and core APIs must reliably deliver on the promises made by the automated assistant.
Van Staden says that retailers should first define the required business outcome before rushing to find a use case for AI. “Then, retailers need to identify the necessary supporting data, validate its quality, and establish clear data ownership and governance rules before introducing AI into the environment. Otherwise, you are just building a fancy, shiny house without a solid, or trustworthy, foundation,” he says.
He says that the stakes of automated basket building extend well beyond standard privacy concerns. The critical task for retailers, he says, is to determine what the AI system is optimising for. “Perhaps it is the lowest cost, highest profit margin, convenience or the customer’s own explicit preferences. Without active and deliberate governance, an algorithm trained on historical purchase patterns may well reinforce past behaviours instead of supporting a customer’s health or budget goals. These are important considerations for all retailers looking to dip their toes into this area,” he says.
Human oversight is critical, he says. This must be designed into the system to ensure recommendations remain sensible, detect potential biases or security risks, enable effective system debugging, and allow operations to adapt to new conditions such as regulatory changes.
“As soon as you remove the human in the loop, you give away ownership and accountability,” he says. “If you do not take accountability, you are no longer in control when something goes wrong or when customer feedback requires a course correction.”
Much like other regions, local retailers need to navigate complex personal information environments and meet strict regulatory requirements. Tokenisation and basic data anonymisation do not eliminate risk on their own because AI models can still group consumers by things such as geolocation, income brackets and purchasing behaviours. “Retailers must give customers meaningful choice and full confidence in how their personal information is gathered and used,” he says.
“The race to deploy AI shopping assistants is accelerating. This is a global phenomenon,” he says. “Rather than partaking in the race for its own sake, the more important question retailers need to ask alongside a specialist partner is whether they are ready to manage the operational, governance, and customer-trust consequences that follow. The shopping experience may become invisible, but accountability must not.”
