Retail AI Report: From Pilots to Proof

‍

Virtual Try-On Technology Takes Aim at Retail's $850 Billion Returns Crisis

AI startup Catches has launched virtual try-on technology on luxury brand Amiri's website, creating digital customer twins with what they describe as "mirror-like realism." The timing is critical: the National Retail Federation reports that 15.8% of retail sales were returned in 2025, totaling $849.9 billion, with online returns hitting 19.3%. Google is making its own virtual try-on technology accessible directly in product search results starting April 30.

Making sense of it: Returns aren't just a customer service issue—they're a margin killer that disproportionately impacts online retailers. What makes this development significant is the convergence of improved AI visualization quality and major platform distribution (Google search integration). For the first time, virtual try-on is becoming both realistic enough to influence purchase confidence and accessible enough to reach mainstream shoppers at the moment of discovery. Retailers should be pressure-testing these solutions now, because the technology has finally crossed the threshold from "interesting experiment" to "competitive necessity." The brands that reduce online return rates by even 3-5 percentage points will see material margin improvement.

Source: CNBC

‍

M&S Makes Massive Bet: 11,000 Microsoft Copilot Licenses for Store and Support Staff

Marks & Spencer is deploying Microsoft 365 Copilot and Agentic AI tools to 11,000 store managers and support center colleagues, representing one of retail's largest workforce AI implementations. The rollout includes a comprehensive training program and a strategic partnership with Microsoft to embed AI capabilities across operations, promising to make tasks like data access and analysis "quicker and simpler."

Making sense of it: This isn't a pilot—it's a wholesale transformation of how M&S employees work. The scale signals genuine conviction that AI assistance will deliver returns worth the substantial licensing investment (likely $30-40 per user per month, or roughly $5 million annually). What's telling is the focus on store managers and support staff rather than just corporate roles. M&S is betting that empowering frontline workers with AI-powered data access and analysis will unlock operational improvements that centralized analytics teams can't achieve alone. For other retailers, the question isn't whether to deploy workforce AI, but how quickly you can move while your competitors are still in committee meetings discussing governance frameworks.

Source: Retail Technology Innovation Hub

Macy's AI Shopping Assistant Users Spend 400% More—But the Industry Wants Proof, Not Promises

Macy's revealed that early users of its "Ask Macy's" AI shopping assistant spend 400% more than non-users, while David's Bridal reported that its AI wedding planning platform Pearl has increased time on site. Yet the broader narrative at retail conferences has shifted decisively from AI experimentation to demanding measurable ROI and comprehensive AI strategies.

Making sense of it: That 400% figure sounds extraordinary—until you ask the harder questions. Are these early adopters already your highest-intent customers? Are we measuring causation or just correlation? The industry's healthy skepticism reflects a necessary maturation. Smart executives know that self-selected early AI users aren't representative of your full customer base, and impressive engagement metrics don't automatically translate to profitable growth. What matters now is moving beyond cherry-picked success metrics to understand full-funnel economics: acquisition cost, conversion rate, average order value, return rate, and lifetime value for AI-assisted versus traditional shopping journeys. The retailers who rigorously measure these will know whether to scale or pivot while competitors are still celebrating vanity metrics.

Source: Modern Retail

Beauty Industry's AI Transformation: NielsenIQ Reports 49% of Consumers Already Using Generative AI for Product Recommendations

NielsenIQ's State of Beauty 2026 report reveals global beauty sales grew 10% year-over-year, with e-commerce expanding six times faster than in-store. More significantly: 49% of consumers already receive beauty recommendations from generative AI, 53% purchase through social platforms, 22% buy directly via TikTok Shop, and in China, livestreaming accounts for 70% of beauty sales on platforms like Douyin.

Making sense of it: Beauty is the canary in the retail coal mine for AI-powered discovery. When half of consumers are already getting product recommendations from AI—not from brands' proprietary tools, but from ChatGPT, Google's Gemini, or social platforms' algorithms—the traditional path to purchase has fundamentally fractured. Brands can no longer assume customers will come to their website or even to traditional search. You need to be discoverable wherever AI agents and social algorithms are making recommendations, which means investing in content and data infrastructure that feeds these systems. The China livestreaming stat (70%) shows where social commerce can go when platforms, content, and transactions fully converge. Western retailers have maybe 18-24 months before similar patterns dominate their categories.

Source: Business Wire

Tesco and Mistral AI Launch Joint Lab, Signaling European Retailers' Turn Toward Specialized AI Partnerships

UK grocer Tesco announced a partnership with French AI company Mistral AI, gaining full access to Mistral's commercial AI models and establishing a joint "AI lab." The collaboration aims to improve customer service, support staff workflows, accelerate data analysis, and create practical generative AI solutions specifically tailored to retail operations.

Making sense of it: This partnership represents an important strategic alternative to the "Big Tech or build it yourself" dilemma. By partnering with Mistral—a European AI specialist—Tesco gets access to frontier AI capabilities while maintaining more control over customization and data sovereignty than typical hyperscaler arrangements would allow. The "joint AI lab" structure suggests genuine co-development rather than just licensing, meaning the solutions will be purpose-built for retail operations rather than adapted from generic enterprise AI tools. For retailers frustrated by the limitations of out-of-the-box AI solutions but lacking the resources to build proprietary models, this middle path of deep partnerships with specialized AI providers may become the dominant model. Watch for more European retailers to follow this playbook.

Source: Crescendo AI

What to Watch

As virtual try-on technology becomes embedded in Google search results later this month, monitor how it affects your product page traffic and conversion patterns—the discovery journey is changing faster than most attribution models can track. And keep an eye on how workforce AI deployments like M&S's Copilot rollout impact operational metrics over the next two quarters; the data will separate transformational investments from expensive distractions.

‍