Instawork’s New Wearable Device Uses Workers to Train Robots

Instawork is planning to launch a new wearable camera system, which it calls an “Instacore.” The device is made up of five cameras mounted on the head, chest, and wrists, which connect to a computer backpack.

Two of the head cameras calculate depth, the chest camera captures the broader environment, and wrist cameras track hand movement. The Instacore’s wrist cameras also have tracking markers that calculate the exact 3D position and angle of each hand, which can be translated into robot joint angle training data.

The device weighs about 3 pounds. Although that isn’t much, Aaron Kerchner, an Instawork Pro and on-site supervisor at the company’s Mountain View facility, told Business Insider that “you definitely have to get used to it.”

Instawork wants workers to wear the Instacore to collect real-world data that can be used to train machines. The company wants to film them performing tasks for their jobs, like chopping vegetables in a kitchen or stacking shelves in a grocery store.

This data will be used to help train robotic systems and AI models to perform tasks in commercial settings. Instawork has about 10 million workers (known as “Pros”) on its platform. The company believes that leveraging them for data will be invaluable to AI labs and robotics companies.

“We will be rolling out Instacore across the world and collecting what we think will be the largest and most diverse commercial data set for robotics,” Aaron Bromberg, head of robotics at Instawork, told Business Insider. 1

Why companies need real-world data for robotics training

Robotics has advanced a lot over the past decade, but these systems still need human data to improve. Like LLMs, which need vast amounts of text and visual data, robotics systems need real examples of how humans move.

One of the biggest challenges in the robotics industry is replicating tactile human responses. For instance, a humanoid robot may struggle with simple actions like chopping vegetables or folding clothes.

Many robotics companies have tried to solve these issues by running simulations for the robots’ AI models to practice in. However, it’s tricky to model things like friction or elasticity in simulations, so the robots weren’t able to replicate what they learned in the physical world.

With that in mind, many robotics firms want to use humans to gather important tactile data. While doing so may be more complex than just running simulations, many researchers believe it’ll make for more efficient training.

One robotics and AI-training startup from San Francisco called Encord has begun using humans to train robots for this exact reason. Fernando Flores, senior robotic pilot at the Californian startup, told The LA Times he’s spent hours using robotic arms to pour cups of coffee, having to start again if there are too many spills.

“Typically, we don’t want any mistakes,” he said. “If we have more than three consecutive mistakes within a 15‑second window, that’s not going to be good data.”

With an ever-growing need for reliable training data, it makes sense that robotics companies are turning to pros within the gig-work space. By turning to Instawork’s portfolio of Pros, robotics companies have access to a huge amount of data.

Instawork’s plan to turn workers into RoboCop-esque data collectors is part of a growing trend of companies looking to move AI from servers into the “real” world.

OpenAI CEO Sam Altman wrote on X that they’ve been ramping up their own robotics efforts, hoping that AI can “help people in the physical world.”

“In the short term, we are focused on robots to support skilled workers to build our future infrastructure; in the long term, we imagine everyone having a personal robot doing anything they need,” he wrote.

In the gig app space, other companies are piloting similar initiatives to Instawork.

A startup called “shift” went viral earlier this month after claiming in their launch video that they’ll clean New York City residents’ homes for free. The catch is that the cleaners will wear head-mounted cameras, with the footage being used to train AI. Harry Kilberg, shift’s General Manager, told Business Insider they had “thousands and thousands of people trying to book” since the launch video went live.

Freelancers on many popular gig apps have struggled in recent years. Workers on apps like DoorDash, TaskRabbit, and Uber claim that the platforms have become oversaturated, making it challenging to find work.

On top of the challenge of finding gigs, there’s also the threat of AI displacement. A SideHustles.com survey found that one-third of freelancers (35%) have lost work to clients who now use AI to complete it.

With that in mind, do AI and robotics training present new earning opportunities for struggling gig workers?

Instawork’s CEO Sumir Meghani seems to think so. He told Business Insider that the platform’s foray into robotics would create new jobs. 1

“Because of the acceleration of physical AI, there are all these new jobs being created, and our focus is on making our Pros relevant in a world with robots,” he said.

Instawork hasn’t clarified whether it’ll pay its Pros more if they participate in robotics training.

While Meghani believes that the Instacore will create more earning opportunities for workers, using them to gather training data for AI may not sit well with every freelancer.

Data ownership and privacy issues

Instawork claims that using the Instacore will be optional and stated that any data gathered will be anonymous.

However, currently, there are no federal guardrails in place when it comes to employee tracking. If a worker does opt to use an Instacore, there’s not much they can do to protect their data afterward.

Side Hustles spoke with Alan Heimlich, President of the San Jose firm Heimlich Law, PC, which frequently provides consulting services for Silicon Valley companies on IP strategies, including AI data ownership issues.

Heimlich argued that there “could be some serious concerns” over privacy and data ownership with these devices.

“It seems that users have nominal control over what data they provide and how that data is used,” he said. “The Instacore would likely capture images of the workers themselves, customers at the site, biometric data such as facial expressions, body language, and anyone who happens to be around during the time of data collection.”

Jared Pope, Benefits and Employment Law attorney and CEO of the workplace misconduct management platform Work Shield, also had a similar take. Pope told Side Hustles that the real question was whether workers knew what they were signing up for.

“Most people hear ‘wearable camera’ and think about the device itself,” he said. “What I’d want to know is: how long is the data being kept, and what happens to that information down the road?”

Using the data against workers

Even if Instawork claims that the data gathered is used strictly for robotics training, some workers worry it could still be used against them.

Meta started tracking workers’ mouse movements and keystrokes earlier this year, using the data to train AI systems. Although the company claimed it wasn’t using any of the data to evaluate employees’ performance, thousands of workers were laid off just months after Meta began tracking them.

There’s no proven link between Meta’s new tracking policy and the layoffs, and the company’s defenders pointed out it had already laid off a large number of employees earlier this year before they began the tracking initiative. But for many workers, the timing of the latest wave of layoffs felt too convenient.

One anonymous Meta employee told the BBC that having all their actions scanned to train AI models amidst a slew of layoffs feels “very dystopian.”

Pope told Side Hustles that monitoring workers has the potential to erode trust between companies and their contractors, claiming that “people act differently when they feel like they’re constantly being watched.”

“If workers feel every interaction, conversation, or decision is being captured and analyzed, it can affect trust and change behavior in ways companies may not anticipate,” he said. “Technology isn’t the problem. A lack of transparency is.”

Companies claim that implementing AI into their workflows will keep human workers relevant in the age of automation. Many workers believe the opposite, and see the technology as the beginning of the end of work as we know it.

The reality is less black and white.

New jobs in an ever-expanding market

The AI-training market is exploding in the US. A report from Grand View Research valued the global data collection and labeling market size at $3.77 billion in 2024. They predict that it will be valued at $17.10 billion by 2030. 2

With more companies exploring robotics, the demand for real human data will only increase. That could mean more jobs for gig workers who are losing out to AI.

Many of the go-to jobs in the gig-work space are already pivoting to AI training. Take data entry; jobs for manual data entry keyers have been drying up in recent years due to AI.

However, a lot of firms now hire independent contractors to review, annotate, and edit AI-generated work. This entire class of jobs was created in the last few years.

DoorDash’s “Tasks” app also pays its workers (known as “Dashers”) to help train AI models. Like Instawork’s Pros, Dashers can get paid to film themselves completing tasks like folding laundry or taking pictures of store shelves to see what’s out of stock. This data is all used to train AI.

None of these jobs would exist if not for the need for AI-training data.

Training your replacement

While AI could open up new opportunities to gig workers, those gains may only be temporary. When companies finally have sufficient data to send robots into the wild, what happens to the workers who were training them?

Erin McGoff, who founded and runs the career-education platform AdviceWithErin, told Business Insider that she constantly hears these concerns from workers. “I have people who say that they, every day, feel like they’re training their replacement,” she said.

It’s hard to dismiss those fears as unfounded. A marketing writer and content manager using the moniker Joe told The Guardian that he “wished he’d seen the writing on the wall” when the company he worked for asked him to create “AI process workflows” and “best practices” documentation.

In August 2025, two weeks after he completed and handed in his documentation, Joe was laid off.

“At my exit interview, I was told it had absolutely nothing to do with my work or performance; they blamed ‘market conditions,’ and some of that was no doubt true, but the timing of it was certainly suspicious,” he said. “Working for this company and being asked to do this—training your robot replacement—feels like digging your own digital grave.”