The most important AI in retail isn’t the part you see. According to MIT Tech Review, the real transformation at legacy retailers is happening behind the scenes: how products surface in search, how inventory moves, how engineers ship code, and how stores react to customer behavior in real time. The flashy stuff, virtual try-ons and chatbot assistants, gets the headlines. The plumbing is where the money is.
Macy’s is the case study. Senior director of engineering Murali Murugan calls the approach “AI-first,” and he’s clear about what that does and doesn’t mean. “AI first isn’t about adding intelligence on top,” he told MIT Tech Review. “It’s about redesigning how decisions happen so the business moves faster and every experience feels more relevant by default.”
That distinction matters more than it sounds. Most companies bolt AI onto workflows they already have. Macy’s is rebuilding the workflows around it, embedding intelligence into personalization, search, operational planning, and software development itself.
What’s actually changing
The shift Murugan describes is the one happening across the whole sector: moving from isolated pilots to integrated systems. He frames the goal as compressing “the gap between the signal and the action.” In plain terms, the time between a customer doing something and the business responding to it is shrinking toward zero.
Macy’s didn’t start with a moonshot. They went after narrow, high-impact wins first, search recommendations and customer engagement, where you can measure conversion lifts and reduced friction fast. Those wins built internal trust. “Once we established the quick wins, scaling was a business decision, not a technology debate anymore,” Murugan says.
That sequence is the real lesson here. The hard part of enterprise AI isn’t the model. It’s getting the organization to stop arguing and start funding. Prove value on something small and measurable, and the political fight disappears.
Why it matters now
Three forces are colliding for legacy retailers:
- Margins are thin and competition is brutal. Amazon and a wave of direct-to-consumer brands have trained shoppers to expect instant, personalized everything.
- AI infrastructure finally got cheap and good enough to run in production, not just in a lab demo.
- The early adopters are pulling ahead. Once one major retailer compresses its signal-to-action loop, the rest are reacting to a faster competitor.
Macy’s is now extending this into conversational commerce with Ask Macy’s, a shopping assistant built to act more like a stylist than a search bar. Describe what you need for a prom, a vacation, or a last-minute event, and it pulls recommendations shaped by your past purchases and context. The interesting part isn’t the chatbot. It’s that the chatbot only works because the underlying data and decision systems were already rebuilt.
The human-in-the-loop bet
Worth noting: Macy’s frames AI as an “invisible layer” that augments human judgment, not a replacement for it. The long-term vision is retail that feels seamless and adaptive, powered by systems customers never notice are running. That’s a deliberate position, and a defensible one. The retailers getting burned right now are the ones who handed customer-facing decisions fully to a model and called it done.
One caution for readers: this piece was produced by MIT Tech Review’s custom content arm in partnership with Infosys, not its editorial staff. So treat it as an informed vendor-adjacent view of where retail AI is heading, not independent reporting. The strategy still holds up on its own logic.
What to do about it
If you’re running AI inside any operations-heavy business, steal the playbook:
- Start where you can measure. Pick a use case with a clear conversion or cost metric. Win there first.
- Rebuild the workflow, don’t decorate it. Bolting AI onto a broken process just makes the broken process faster.
- Keep humans in the loop on customer-facing calls. Augment judgment, don’t outsource it.
- Treat scaling as a budget decision, not a tech debate. If you’re still arguing about whether AI works, you haven’t shipped a real win yet.
The next 12 to 24 months will separate retailers who quietly rewired their decision systems from those still running pilots. As Murugan puts it, the real edge comes from “continuous improvement,” the boring compounding of timing and execution. Boring usually wins. Full details are at the original MIT Tech Review piece.