The robots are coming. You already knew that.

What you probably didn't know: the workers training them aren't getting paid for it. Their hand movements, their muscle memory, their years of skill. Filmed. Packaged. Sold to the companies building the machines that will replace them.

The article is below. It's worth your time.

Read. 👇

Sponsored by:

The workforce is changing faster than most recruiting teams can track. Automation is rewriting job descriptions, displacing roles, and creating demand for talent that didn't exist three years ago.

ProvenBase finds the candidates your competitors aren't calling. Passive talent. People already employed, not actively looking, not on the usual boards. The ones worth pursuing.

While the robots are being trained, the humans who manage them still need to be hired.

Find your next human hire at ProvenBase.com

Inside Every Robot are Workers Who Never Got Paid

Two things happened on June 24, 2026. On one side of the world, factory workers in India were fitted with head-mounted cameras. Their hand movements, their stitching rhythms, their years of bodily instinct. Filmed, annotated, packaged. On the other side, Agility Robotics went public in a $2.5 billion SPAC deal. Its humanoid robot, Digit, is already walking Amazon warehouses. Agility now has $420 million to build 10,000 units a year.

Same pipeline. Two ends. One captures the raw material. The other ships the product.

The workers in Gurugram and the investors on Wall Street are inside the same machine. The outcomes are not the same.

The Problem Robots Cannot Solve Alone 🤖

ChatGPT learned from text. Billions of words, already online, already scraped.

Humanoid robots have no such shortcut. They need to move through the physical world. Grasp a shirt. Stack a box. Lay a brick. That knowledge does not live in any database. It lives in human bodies.

The answer the industry landed on: egocentric data. First-person video, shot from head-mounted cameras and smart glasses, capturing exactly how skilled hands do skilled things. The micro-corrections. The spatial judgment. The rhythm a mason builds after ten thousand bricks.

Experts now describe data as the biggest bottleneck in robotics. The estimates are staggering. Hundreds of millions of hours of human activity. Maybe billions. All of it filmed, cleaned, and fed into training pipelines before robots can operate reliably at scale.

India as the World's Data Factory 🌏

India already owns roughly 35% of the global data annotation market. About 60% of that revenue flows from US clients. The math is simple: data collection that runs $30 per hour in America costs under $5 in India.

EgoLab, Humyn AI, FPV Labs, Micro1, Egodata, Scale AI, Objectways. All of them have built industrial-scale pipelines pulling footage from garment factories, construction sites, and warehouses across South Asia.

The workers generating this data are often getting nothing extra. Some earn $200 a month to stitch shirts. They are told to wear cameras. Occasionally they receive a soft drink. The companies argue that factory management is compensated for facilitating access. Critics call that a deliberate misdirection. The hands doing the work are not the hands getting paid.

The US Manufacturing Squeeze 🏭

Two forces are compressing American manufacturing at once.

First: demographics. The workforce is aging. Factories are bleeding experienced workers to retirement, and the pipeline of replacements is thin.

Second: reshoring politics. Federal pressure to bring production back home requires either domestic labor or its robotic substitute. Labor is expensive and scarce. Robots are getting cheaper and smarter.

Agility CEO Peggy Johnson named both directly as tailwinds for Digit sales. Foxconn, Amazon, Nvidia, and SoftBank backed the company's $2.5 billion valuation. That money says the demand is structural.

The Investment Surge 💰

The Global Humanoid Robot Market sat at $1.49 billion in 2023. It is projected to hit $29.12 billion by 2033, growing at 34.62% annually. The big players are not waiting:

• Agility Robotics: $2.5 billion valuation. Digit robots at Amazon and Toyota.

• Tesla: The Optimus robot. Elon Musk claims it will eventually represent 80% of Tesla's total value.

• Figure AI: Partnering with BMW and US Steel, backed by over $2.6 billion.

• Boston Dynamics: Atlas robots, deployed with Hyundai.

EgoLab, the Indian data aggregation company that collected footage from the Gurugram factories, claims Tesla, Boston Dynamics, and Figure AI among its clients — a characterization reported directly by The Guardian as well. The data flowing out of South Asian factory floors is the raw fuel behind these valuations."

The Automation Clock ⏱️

Where We Are Now: 2024–2026

Digit robots move and stack heavy containers in Amazon facilities. Optimus units run simple assembly at Tesla. Both are capable specialists inside narrow, structured environments.

The gap is dexterous manipulation: handling small, irregular, unpredictable objects. That is exactly the gap egocentric data is designed to close.

The Automate Show 2026 Blog marks this year as the pivot. Industrial humanoid robots are moving from experimental prototypes to real-world pilot fleets.

The Near-Term Horizon: 2026–2030

Analysts expect commercial-scale deployment across automotive, electronics, and logistics before 2030. Automotive has historically led industrial automation, representing 38% of existing industrial robots.

A landmark MIT Sloan study found that each additional robot installed in a US commuting zone eliminates about 6.2 workers and cuts wages by 0.7%. The impact is nearly twice as severe in lower-skilled regions. The communities most reliant on manufacturing absorb the hardest hits.

The Medium-Term Horizon: 2030–2040

Oxford Economics projected up to 20 million factory jobs displaced globally by 2030. The speed depends on data. More egocentric footage means faster capability expansion, reaching the messy, variable conditions of real factory floors sooner.

Who Gets Paid for the Body's Knowledge? 💸

The Disconnect

A garment worker paid $200 a month is being paid for shirts. When she also films her movements to train a robot, she is simultaneously producing a second asset. A dataset. Durable, scalable, licensable, permanent.

Ohio State University Associate Professor Madhumita Dutta frames it plainly: workers are selling labor for wages, but here they are also generating a commercial digital asset. Without awareness, there is no negotiation.

The Emerging Framework: Intellectual Equity

A growing argument in academic and policy circles calls what workers produce "intellectual equity." Their contributions to AI systems are long-term assets. They should generate recurring compensation, the way a musician earns royalties.

MIT Sloan professor Thomas Malone has proposed "learnrights": an exclusive right for data contributors to license their content for AI training, negotiated for fair payment. Legal scholars writing in the Virginia Law Review have pushed for consent and compensation mechanisms to resolve the generative AI copyright crisis.

Models on the table:

• Royalties: Recurring micropayments tied to ongoing AI usage.

• Inference-time compensation: Payment triggered when AI applies patterns from a specific worker's data.

• Data unions: Workers pool data and negotiate licensing fees collectively.

• Equity: Workers receive minority stakes in the AI companies their bodies trained.

The Legislative Gap

The US has a bill. Senate Bill S.2367, the AI Accountability and Personal Data Protection Act, introduced in July 2025. It would give individuals a federal cause of action when their data gets used in AI training without express, prior consent. As of mid-2026, it is stalled in the Senate Judiciary Committee.

The EU has GDPR and the AI Act. Consent requirements, transparency obligations. Neither pays workers for the physical data leaving their bodies.

India has a law too. The Digital Personal Data Protection Act, 2023, with rules notified in November 2025. It has a broad employment exemption. No biometric protections. No collective enforcement. Researchers confirm what the text already telegraphs: egocentric factory data is ungoverned in practice.

The footage walks out of the country. The workers stay. Nothing follows them home.

What This Means for You 🎯

For Job Seekers

The immediate concern is not just displacement. It is that your skill is being used to train the machine that will replace you, without your knowledge or consent.

Repetitive assembly, material handling, basic sewing, sorting, stacking. High risk. Complex social judgment, environmental unpredictability, emotional intelligence. Far more resilient.

The traditional entry-level pathway into skilled manufacturing work is compressing. The first rung is being automated out.

For Recruiters

Demand for traditional manufacturing roles will contract in volume but shift in profile. Manual capacity is out. Technical oversight is in. Robotics engineering, AI systems integration, human-robot collaboration management. The roles are growing. The candidate pool is not keeping pace.

Recruiters also carry a new ethical weight. If a client is collecting egocentric data from its workers, staffing those roles may raise questions about informed consent, data rights disclosures, and surveillance exposure. Know what you are placing candidates into.

For HR Leaders

A Boston Consulting Group analysis found AI will reshape 50 to 55% of US jobs over the next few years, with 10 to 15% potentially eliminated. HR must lead massive, continuous reskilling programs. Not optional. Not gradual. Urgent.

The Guardian investigation found that in some facilities, egocentric footage was used not only for AI training but to generate productivity reports and track social interactions. That dual use creates legal exposure under employee monitoring laws, as Dykema legal analysts note.

HR Reporter captures what is not easily quantifiable: the psychological impact of anthropomorphized machines on the factory floor. Job insecurity is one thing. Working next to something shaped like you is another. HR has to manage both.

For Workforce Planners and Policymakers

The burden will fall hardest on communities already economically fragile. Industrial heartland. Lower-skilled regions. The same places the MIT Sloan data shows absorbing twice the wage and employment damage.

Retraining funds, labor protections, structural economic reform. Forbes covers how AI displacement has reignited serious conversations about Universal Basic Income (UBI) and robot taxes to fund social safety nets.

But the most under addressed intervention is data rights legislation. The value flowing from egocentric data is enormous. Right now it is captured almost entirely by technology companies and their investors. Establishing legal frameworks for data ownership, consent, and compensation is not a future policy debate. It is overdue.

The Body Walks Out. The Data Stays. ☠️

"Sarayu Natarajan, founder of the Bengaluru-based Aapti Institute, put it precisely: 'The data originates in a worker's body and actions, but once extracted it no longer remains attached to them in the same way.”

Think about what that means. A garment worker's hand movements get filmed. The footage trains a robot. The robot learns to sew. And eventually, the robot does not need her.

The company keeps the robot. It keeps the dataset. It keeps the intellectual value of her lifetime of work. She got a soft drink.

The capital is committed. The data pipelines are running. The robots are already on the floor. What is not in place is any framework to ensure this transformation is just.

Lalita, the garment worker at the center of The Guardian's investigation, asked the question plainly: "Who is going to pay us when we are replaced by robots?"

Nobody has answered her.

The HR Blotter

The Hospital Didn't Lose the Nurses. It Ran Them Off. - The National Council of State Boards of Nursing reports 100,000 registered nurses left during the pandemic, and another 610,388 plan to follow by 2027, including nearly 189,000 under the age of 40. Healthcare keeps calling it a pipeline problem. The nurses already in the system, skipping lunch, covering vacancies, absorbing chaos, and driving home wondering if the degree was a mistake, are calling it something else. You cannot recruit your way out of a job people are trying to survive.

North Carolina Paid $47 Million in Fraudulent Unemployment. Roy Cooper Is Now Running for Senate. - A North Carolina State Auditor report found $47.2 million in fraudulent unemployment insurance payments made during and shortly after Roy Cooper's tenure as governor, with only $12.1 million recovered. The improper payment rate averaged 22 percent, more than double the federal threshold, and Cooper's administration received a $6.8 million federal grant to fix the problem, then took three years to do anything with it. The state auditor called his lack of urgency "concerning." Cooper is currently asking North Carolinians to send him to Washington.

The AI Gold Rush Has a Pets.com Problem - An MIT study found that 95% of companies investing in generative AI have yet to see measurable returns, AI tool adoption among large US firms dropped from 14% to under 12% in just two months, and OpenAI is reportedly seeking $1.5 trillion in computing capacity against revenue streams that don't come close to justifying it. Nvidia hit $5 trillion in market value while nine of the ten most valuable companies on earth are now AI stocks, accounting for roughly a third of the entire S&P 500. We have seen this movie before, and Pets.com did not have a sequel.

Two-Thirds of HR Leaders Don't Know What AI Can Do. The Other Third Mostly Isn't Compliant. - SHRM's State of AI in HR 2026 found that 67% of HR leaders cite lack of awareness as their biggest barrier to AI adoption, and 57% of HR professionals in AI-regulated US states don't know the policies governing their own hiring tools. Among those who do know the rules, only 12% have implemented compliant practices. Ninety-two percent of CHROs want more AI in their organizations. The people who will actually run it have no idea how it works.

The Entry-Level Job No Longer Exists. It Was Replaced by a Job That Used to Take a Decade to Earn. - A PwC study found that AI-exposed entry-level roles are now seven times more likely to require traditionally senior-level skills like judgment and leadership, early-career AI job postings have flatlined, and listings for junior roles demanding mid-career or senior experience have grown 35% since 2019. AI-driven job cuts hit 87,174 through May 2026, already outpacing all of 2025. Companies stopped building careers from the bottom up, and nobody told the people at the bottom.

…

The Jim Stroud Podcast

Not subscribed to The Jim Stroud Podcast? Then you’ve been flying blind. Here’s a sneak peek at the latest episode debuting tomorrow.

…

…

One more thing before I go…

The question Lalita asked is still unanswered.

She is a garment worker in India. She wore a camera to work. Her hands trained a robot. Nobody told her that was happening. Nobody paid her for it.

Somewhere in a warehouse right now, that robot is working her shift.

Policymakers are stalled. Legislation is sitting in committee. The data pipelines are not waiting.

The workforce intelligence in this newsletter exists for one reason: so you see what is coming before it arrives at your door. Share it with someone who needs to read it.

See you next week.

…