Macy’s is betting that AI works best when customers never notice it. Senior director of engineering Murali Murugan describes an “AI-first” redesign of core retail operations, including search ranking, inventory planning, and developer tooling, built in partnership with Infosys and covered in this MIT Technology Review webcast<\/a>. The centerpiece is Ask Macy’s, a conversational shopping assistant that draws on purchase history and stated context to surface styled recommendations. The broader play is compressing the lag between a customer signal and a business response.<\/p>\n Retailers who’ve spent the last two years running AI pilots in isolation are now watching peers like Macy’s declare those pilots over and fold the results into production systems. The question for any CMO with a personalization budget isn’t whether to follow that path, it’s whether your data infrastructure, meaning the purchase history, behavioral signals, and preference data that feed a tool like Ask Macy’s, is actually clean enough to make the move. A conversational interface built on stale or siloed data doesn’t become a stylist. It becomes a more confident version of bad search.<\/p>\n The piece, produced in partnership with Infosys (which sells the implementation services to build exactly this kind of stack), frames the Macy’s story as a repeatable playbook: start with narrow wins in search and engagement, build internal momentum, then scale. That framing flatters a managed-services model where the vendor stays involved at every stage. The underlying sequencing is still sound, but the timeline compression Murugan describes, from pilot to integrated operating philosophy, is the part that tends to run longer and cost more than a vendor-adjacent case study suggests.<\/p>\n The leading indicator to watch is whether Ask Macy’s drives measurable lift in average order value or conversion on assisted sessions versus unassisted ones. If the conversational layer is genuinely acting like a stylist, those numbers move together. If it’s just repackaging keyword search in a chat wrapper, conversion holds flat and return rates climb because recommendations miss on fit. That distinction is the budget decision hiding inside this story, not whether to invest in AI-assisted commerce, but whether your current data layer can support the version that actually works.<\/p>\nWhat this means for your business<\/h2>\n