Somewhere in a stockroom or a restaurant kitchen right now, someone is working an ordinary shift with a recording device strapped to their head that they didn't bring from home. A company mailed it to them.
That's the actual mechanics behind Index, the platform Figure AI launched on August to pay people for filming themselves doing everyday work, cooking, folding laundry, stocking a shelf, and working a warehouse aisle. It’s all done so its humanoid robot, Helix, can learn from what they capture.
In its first weeks, the app had been downloaded more than 264,000 times across 108 countries. What Figure AI is calling “creators” have uploaded 16 million videos, nearly five years of human work, every single day, according to the company’s own numbers.
Figure has already paid out $15 million to creators, and says it has committed $1 billion to data collection and compute over the next 12 months. What that number doesn't touch is the room itself, the actual physical problem of turning a stockroom or a kitchen into a place a robot can learn from reliably, calibrated, synchronized, lit consistently, on a schedule a real business can live with. That's a production problem, and it's one AV integrators already know how to solve.
Two Tiers, One Gap Nobody's Filling
Index runs on two distinct tiers. The first is casual - anyone can become a creator and record themselves doing real tasks in their own home or workplace, on whatever device they already own. Across this gig category more broadly, that has meant exactly what it sounds like.
MIT Technology Review reported on one worker, a medical student in Nigeria training humanoid-robot models after his hospital shifts, who turns on a ring light, straps his iPhone to his forehead, and records himself making a bed, raising his hands "like a sleepwalker." This casual tier is what's driving Index's headline numbers, and it reads a lot closer to gig work than production work.
The second tier is different in kind, not just scale. A household or a business can book a creator to come help with chores or errands, "at home or across business areas like logistics, restaurants, factories, and offices," and accepted creators working this tier "sign up to receive a recording device," rather than filming on a personal phone.
Independent coverage of the app shows what that device looks like in practice, head-worn capture gear rather than a handheld camera, though Figure hasn't published the technical specification publicly.
It's a meaningfully different job than the one most coverage describes. A casual creator is filming their own kitchen with a phone they already own. A paid creator is showing up at someone else's logistics center or restaurant, wearing equipment the company supplied, expected to produce usable, first-person footage of a real business in operation. Only one of those two jobs looks anything like gig work once it's actually happening.
The Same Problem AV Already Solved
Strip away the robotics framing, and the paid tier is a calibrated, multi-sensor, on-location capture job - getting clean, synchronized, first-person footage out of a working business without disrupting the business itself. That problem isn't new. It's the exact discipline AV integrators already practice for broadcast production and volumetric capture, sync, lighting consistency, network throughput, storage, solved for environments that were never built to be filmed.
Run that kind of shoot without an integrator's oversight and the failure modes are the same ones broadcast crews spend careers designing around - audio drifting out of sync with the footage, lighting that shifts from one take to the next, a wireless link that drops mid-shift in a warehouse full of metal shelving.
The difference is who's doing the solving. A broadcast client hires a systems integrator because they already know that's the job. A restaurant or a warehouse that books a creator through Figure's app almost certainly doesn't think of what's about to happen as a production shoot, because nothing in the booking flow tells them it is one. The room becomes a set for an afternoon, and the business that owns it never gets asked whether it needs the kind of planning a set usually gets.
Elsewhere in this same data-collection economy, pay for comparable recording work has been reported as low as $2.60 an hour, a rate nobody would offer a production crew. Figure is quietly re-inventing calibrated, first-person, multi-sensor capture from first principles, one paid creator visit at a time, for a customer, physical-AI companies buying training data at industrial scale, that doesn't yet know an AV integrator is the person who already does this for a living, or what that work is actually worth.
Figure Isn't Alone
The same underlying problem is showing up on the industrial side too, just solved differently.
On 16 March 2026, Universal Robots and Scale AI launched the UR AI Trainer, a system that captures robot training data through what the companies call a leader-follower setup. A person guides one robot, the "leader," through a task by hand, while a paired "follower" robot copies the motion instantly, and the whole exchange, movement, force and what the camera sees, gets recorded at once.
Anders Beck, Universal Robots' VP of AI Robotics Products, and Ben Levin, Scale AI's General Manager for Physical AI, announced the partnership at Nvidia's GTC conference, with a large industrial dataset built on UR robots due later this year.
It's a very different capture method than a creator wearing a company-issued rig in a restaurant kitchen, but it rests on the same underlying bet, that physical-AI companies need enormous volumes of real-world, sensor-rich demonstration data, and each one is building its own bespoke way to capture it rather than treating it as a solved problem.
Between Figure's consumer-facing gig model and Universal Robots' industrial-grade rig, the middle ground, a reliable, repeatable way to calibrate an ordinary commercial space for that kind of capture, is still nobody's specialty.
Figure's $1 billion, 12-month commitment to data collection isn't a one-time launch. It's a scaling plan, and scaling a paid, on-location, calibrated-capture business means booking more creators into more restaurants, warehouses and stockrooms every month for the next year.
The Capture Specialists Are Already Moving In
Some capture specialists have already made the leap Figure hasn't asked anyone to make.
Apple Arts Studios, a motion-capture house with more than a decade in film and animation, now sells the same optical rigs it built for movies as training datasets for humanoid-robot clients: multiple performers tracked at once inside a calibrated volume, sub-millimeter precision, delivered in the specific file formats robot platforms need. Neither that studio nor its competitors are AV integrators.
They're proof that the skill Figure's paid tier is missing, calibrated, synchronized, multi-camera capture at production quality, already has a home industry, and robotics companies are paying to import it rather than build it themselves.
That's the concrete version of the question this piece keeps circling toward. AV integrators already have most of what a studio like that is selling–broadcast-grade calibration, multi-camera sync, on-location production discipline–just built for corporate AV and events instead of a motion-capture volume.
Physical-AI companies are going to keep needing this kind of capture at scale; Figure's billion-dollar, twelve-month commitment guarantees that much on its own. The question isn't whether that market exists. It's whether AV integrators go after it before it's gone, because Ropedia just raised $22 million to build a purpose-made capture rig for exactly this job, and it isn't waiting for anyone in AV to bid on the work.