According to McKinsey, agentic AI applications and process automation could free up to 40 percent of the time retail employees currently spend on repetitive tasks. Retail leaders and managers could redirect that time toward strategic activities — for example discovering new products, understanding customers better, and optimizing supply chains.
Despite that potential, McKinsey’s market research found that more than two-thirds of retailers say AI tools have had minimal or no impact on their companies so far. Nearly the same proportion reported that their organizations are either unprepared or only minimally prepared to adopt AI. Behind these answers lie structural obstacles: many firms have fragmented IT systems and messy databases that hinder effective AI deployment.
How it looks in practice: US progress and category managers
In the United States, 39 percent of retail companies have already taken initial steps toward adopting AI. One concrete effect is visible in the work of product category managers — those responsible for the assortment, procurement, pricing and promotion of dairy, meat, cleaning products, apparel, electronics and other categories — who now often collaborate with AI agents.
That collaboration requires new skills: category managers need to learn how to prompt the AI models they use effectively and to understand the associated risks, particularly model hallucination, where the system returns incorrect or fabricated information.
Training, readiness and the outlook
In McKinsey’s survey, 24 percent of executives reported that their companies are delivering moderate or higher levels of AI training. McKinsey argues that AI could enable retailers to become more like old‑style shopkeepers, by offloading much of the background work involved in managing large assortments and customer bases so owners and managers can focus on customer relationships and merchandising.
Overall, McKinsey’s findings highlight a significant efficiency upside for retail from AI, but also that substantial technical and organizational barriers remain before that potential is realized at scale.
