AI assistants in retail: How they support staff at the point of sale

A busy store, long queues and a customer with a very specific product question: in brick-and-mortar retail, advice often has to be given quickly – yet the right information is not always readily at hand. With large product ranges in particular, even experienced teams reach their limits.

At the same time, customers expect the same speed and precision in store that they are used to from online shopping. AI assistants are designed to close this gap by providing employees with the knowledge they need exactly when they need it.

DIY stores often stock tens of thousands of items – from small parts and tools to specialist accessories. Even experienced professionals cannot know every detail of such a range. This is where Hey ARMA comes in: an AI-powered tool that the US hardware chain Ace Hardware has rolled out nationwide in more than 2,300 stores. The system runs directly on employees’ smartphones or mobile devices. When customers have specific questions – about product comparisons, instructions or repairs, for example – the AI assistant searches the company’s entire knowledge base and delivers the answer directly at the point of sale. Even if a product was originally bought from another retailer, the tool helps to find a suitable solution quickly. Andy Enright, Senior Vice President, Retail Strategy & Operations at Ace Hardware, describes the approach as follows: “At Ace, we believe technology should make the customer experience better by helping our people do what they do best.”

Technically, the AI assistant runs in real time during the customer conversation – via voice command, text input or live transcript. The personal conversation remains in the foreground, as the AI does not itself interact with customers as a chatbot. Another advantage: new employees, too, are spared time-consuming look-ups in various systems, can provide information from the outset and can concentrate on personal, solution-oriented advice. However, these advantages are offset by certain challenges.

A first critical point is the data. Incorrect or outdated data can lead to wrong information. In addition, there is the risk of so-called AI hallucinations, in which a system generates information that sounds plausible but is factually incorrect. An example: a customer asks whether she can still return a product within 30 days. The AI draws on an old policy and answers yes, even though only 14 days now apply. This is annoying for the customer – and at least as annoying for the company. To prevent this, the knowledge base must be maintained and updated regularly. That takes effort: the AI does not simply work on its own – keeping information current and reliable requires clear responsibilities and processes in the background. In short: An AI is only as good as its data.