GreyOrange has acquired Site Bionics, a developer of AI-powered digital twin software for shops, in a deal announced on 8 October 2026. Financial terms were not disclosed.
GreyOrange, which describes itself as the global leader in AI-powered multi-agent warehouse orchestration and store inventory software, sells gStore, a retail application built on item-level RFID. Site Bionics comes at the same shop floor from the opposite direction, taking the security cameras a retailer already has on the ceiling and turning them into what the company calls a live, continuously updated model of the store and the people moving through it. A digital twin, in that sense, is a software copy of a physical space kept in step with the real thing by a constant feed of live data. Combined with gStore’s item-level RFID, GreyOrange says retailers get a single AI-driven view of both the products on the floor and the people interacting with them.
The founders are the most revealing detail in the announcement. Site Bionics was founded by Adam Blair, the founding CTO of RADAR, and Don Gillett, the former Head Architect of Amazon Go. RADAR is a retail technology company whose own platform pairs item-level RFID with computer vision to report where stock is sitting in a store in real time, so Blair arrives at GreyOrange from a direct competitor in precisely the space this deal is meant to open up. Amazon Go was Amazon’s checkout-free convenience store format, opened to the public in Seattle in January 2018, which used ceiling cameras and shelf sensors to bill shoppers for what they picked up on the way out. The Site Bionics team has joined GreyOrange, and the technology goes into gStore first, with plans to extend it into warehouse operations through GreyMatter, GreyOrange’s warehouse orchestration software.
“Adam and Don have spent their careers at the frontier of store intelligence, from item-level RFID to the camera systems behind Amazon Go,” said Akash Gupta, CEO of GreyOrange. “Together, we’re giving retailers the ability to do more with the infrastructure they’ve already invested in. The cameras on their ceilings and the RFID tags on their products become one AI-driven picture of everything happening in the store. That means fewer blind spots, faster decisions, and the insight needed to drive more in-store conversions.”
What the combined system is claimed to do
GreyOrange sets out four things the merged product is meant to deliver. First, a live digital twin of the whole store with no new hardware, showing how shoppers and staff move, where they linger and what they interact with, off the existing cameras. Second, AI that explains what it sees: video-language models, trained on video and text together so they can answer questions in plain words about footage, would let a head office ask “Which stores set up the new display as planned?” instead of scrubbing through hours of recordings. Third, RFID and computer vision in one model, which GreyOrange says yields an in-store sales funnel down to the individual item and automates loss investigations so asset protection teams get faster, fuller reports. Fourth, one system in place of the patchwork of separate inventory, shopper behaviour and loss tools most retailers run today.
On privacy, the company says all tracking is anonymous, video is processed on site, and the system sits behind enterprise-grade security and privacy controls. Those are GreyOrange’s assurances rather than audited findings, and a UK retailer putting camera analytics into a shop still has its own obligations to work through.
A promise that has gone unmet
“For years, this industry has promised retailers that RFID and computer vision would come together and show them the full story of their stores. That promise has gone unmet,” said Adam Blair, Co-Founder of Site Bionics. “Don and I built Site Bionics to solve the computer vision side of that equation. GreyOrange has mastered the RFID side. Together, we’re finally delivering on it.”
That is a fair description of where the market has been, and it is worth being precise about why the join is hard. The two technologies answer different questions. RFID gives you identity: a read returns a unique number that resolves to one specific unit of one specific SKU, and it does that whether the item is on a shelf, in the stockroom or buried under other stock, because it never has to be seen. Cameras give you behaviour: who reached, who lingered, who walked away, all inferred from appearance. Fusing the two at item level means asserting that the number read off a tag and the object a camera watched someone pick up are the same thing, and that assertion is where these systems have historically come apart.
The failure modes are mundane and stubborn. Items disappear from view into bags, baskets and pockets while their tags keep reading. Tags read through walls and shelving from the wrong fixture, or from the next department along, so the inventory record and the camera’s view of the aisle disagree. Overhead reader arrays give wide, continuous coverage but coarse position, while a handheld gives a clean cycle count only when somebody walks the floor with it, so the twin’s freshness varies by zone and by shift. And a shopper who picks up three garments, carries them to a mirror and puts two back has produced one camera sequence and several tag movements that have to be reconciled correctly, or the sales funnel is wrong at exactly the item level the product is sold on. Hiring the team that built the camera side of Amazon Go is a serious answer to that. It is also why the promise Blair describes has gone unmet for as long as it has, and why what matters next is evidence rather than architecture.
“After years of putting cameras to work across industries, I’ve learned that a camera’s real value is in what it measures, not what it records,” said Don Gillett, Co-Founder and CTO of Site Bionics. “Site Bionics brings that thinking to retail, using the cameras already on every store’s ceiling. Paired with gStore’s RFID data, the store’s digital twin knows not just where people are, but what they’re doing with the products around them.”
What the announcement does not include
There is no price, no location given for Site Bionics, no named Site Bionics customer and no third-party measurement of what the camera platform achieved before the deal. Everything quantified belongs to gStore, and the figures are GreyOrange’s own: better than 99% inventory accuracy, position to within 3 to 5 ft with store maps down to aisle and shelf, a 40% lift in on-floor availability, twice the associate productivity and go-live in under four weeks. GreyOrange also says gStore is the only platform supporting both overhead and handheld RFID in a single store. Neither the announcement nor the gStore product material names the radio frequency band the tags use.
The scale behind the claim is real enough. GreyOrange said in early September 2026 that more than 3,800 retail locations run on its technology, and H&M Group has been rolling gStore out across hundreds of its own stores since around September 2026. That is a substantial installed base of item-level RFID to point a camera model at, and it is why the deal is worth following: if the fusion works anywhere at scale, it will show up first in stores that already have both halves of the equation in place. Until a named retailer publishes numbers, this remains an acquisition release with a good argument inside it.
More store-level deployments are covered in our retail section and platforms of this kind in software. Terms used here are explained in the RFID glossary.
