Why retail’s next AI frontier is Physical, and starts with precision 3D data
By Alex de Vigan, Founder & CEO, Nfinite
Last week, videos of Shenzhen-built humanoid robots stacking breakfast trays lit up LinkedIn feeds. The manufacturer, Dobot, says mass production will begin by mid-2025, at a price tag that undercuts most warehouse cobots today. It was a small vignette, but it signals a big shift: embodied, or ‘Physical’ AI is leaving R&D labs and walking straight toward the shop floor.
For retailers, that walk is a sprint. Robots that can see, grasp, and reason will soon restock end-of-aisle displays, guide click-and-collect runners and sweep dark stores after closing. Yet, the old maxim still applies: garbage in, garbage out. What’s not being discussed is the fact that if the data those systems consume is vague or inconsistent, we can expect phantom stock, mis-picks, and safety incidents to persist. In 2024 alone, inventory distortion siphoned $1.7 trillion from global retailers according to IHL.
“To me, AGI will not be complete without spatial intelligence. And I want to solve that problem.” — Dr Fei-Fei Li, the ‘godmother of AI’. Li’s warning applies just as sharply to retail transformation.
Spatial intelligence begins with precise, photorealistic digital twins, virtual replicas that encode every SKU’s geometry, weight, texture and barcodes. Update a twin once and every downstream system (from warehouse management systems and planogram software to e-commerce viewers and autonomous mobile robots) works from the same single source of 3D truth.
What precision 3D data unlocks
- Scalable content generation: Once a digital twin is created, it can be instantly reused to generate high-quality lifestyle imagery, packshots, videos, and interactive experiences, without ever reshooting a product. This approach enables retailers to visually localize SKUs, test campaigns faster, and personalize content at scale for different markets or customer segments.
- Live stock visibility: Real-time digital twins expose out-of-stocks before shoppers do, cutting lost-sale risk.
- Faster site launches: Plug-and-play 3D assets let new stores or distribution centers come online in weeks, not months.
- Safer, smarter robots: Robots trained on accurate 3D data can navigate crowded aisles without bumping into strollers or shelf displays.
Waymo’s driverless fleet just crossed 10 million paid rides in four U.S. cities. Their secret isn’t only lidar, it’s meticulously curated high-quality 3D data. Retail robots, e-commerce systems, and marketing teams will demand the same diet.
The investment gap no one talks about
Beijing has already launched an $8.2 billion National AI Industry Fund, part of a broader ¥1 trillion push into embodied intelligence. Europe’s dedicated robotics work-programme for 2023-25 totals just €174 million, and the UK’s Innovate budget for advanced robotics this year is £60 m. Europe’s retail heritage is strong, but you can’t run a marathon in flip-flops. Without serious investment, the Physical-AI race may be lost before it begins.
A practical path forward
- Make 3D data a core SKU asset: Retailers should ensure their top-selling products have accurate size, shape and surface detail before feeding them into robots, planograms or 3D e-commerce tools. Data gaps should be fixed before piloting new tech.
- Build a ‘twin pipeline’: Establish a streamlined ‘twin pipeline’ to ensure that product changes (such as new colors or pack sizes) trigger automatic updates to 3D files across all connected systems, from warehouse software to planograms and e-commerce platforms.
- Link store metrics to data quality: Integrate store performance metrics with product data quality by correlating factors like shelf availability or robot collision rates with the accuracy of each digital twin. Use these insights to improve accountability and maintain data integrity across teams.
The aim is to give every picker, planner, robot and shopper the same reliable digital version of every product, so physical and digital operations stay in lock-step.
The call to action
Europe’s retail leaders missed the smartphone supply-chain boom; they can’t afford to miss the Physical AI one. That starts with demanding accurate, shelf-ready 3D data today, before a mis-sized carton jams tomorrow’s picking robot.
Physical AI isn’t a buzzword; it’s a balance-sheet imperative. Retailers who feed their systems pixel-perfect twins will enjoy fuller shelves, happier staff and safer stores. Those that don’t may soon watch a new wave of £21k Shenzhen robots stocking someone else’s inventory.
Alex de Vigan is Founder and CEO of Nfinite, a Paris-based platform developing high-quality precision 3D datasets, powering retailers and AI developers worldwide.
