The future of commerce.
Commerce is where most people meet new technology, and physical AI is about to change it. This track studies two sides of that shift. The first is the selling floor: how embodied AI, from in-store systems and smart fixtures to robots that can show, explain, and hand over a product, changes how goods are sold and how customers experience a store. The second is the workforce. As physical AI enters retail and sales, the people who work in commerce need new skills, and some roles change or move. The question the track is built around is how to develop that workforce so the technology raises what people can do instead of replacing them, and how a business can put physical AI to work on the floor in a way that actually sells.
The track is led by Kat Moore, EdD, MBA, an Industrial Research Fellow hosted in the Glass Lab, drawing on a career in sales and education. The workforce half of the work sits squarely in the lab's own subject: how a frontline workforce is trained for equipment that changes yearly.
↓ White paper · PDFRead onlineTR-2026-31 · research / review preprint
The review states eleven hypotheses about the selling floor and the workforce and tests each against a graded ledger of eighteen deployment records, resolving every one to a position on the trajectory, the constraint currently binding it and the measured change that moves it. The ledger sorts by how much of the change the customer has to absorb: staff-side automation is at chain-wide scale, while formats that ask the customer to change what they do at the exit are optional, licensed into high-throughput venues, or re-sited. The iteration cycle that produces is measured in years, which is what sets the training problem: a curriculum tied to a specific vendor system has a shorter useful life than the competencies that run across all of them. The second half reads the labour series, where the aggregate holds close to flat while the shape of the entry-level role changes.
The finding, made operable.
Ask how often a shelf needs to be checked and the two ways of knowing sort themselves into a fixed sensor that draws continuously and a mobile one that pays per trip. Wait long enough and the answer changes sides. Then the same eighteen-record ledger, grouped by how much of the change the customer has to absorb.
Commerce is where most people first meet a new technology, and embodied AI is arriving on the selling floor ahead of the training that would let the people there work with it. The gap is a curriculum and a credential, both buildable now. That makes this a scheduling problem rather than a research one, and scheduling problems get solved by starting.
One of eight, and only one of them is physics. How we read a frontier →