AI Training Is Creating a Growing Layer of Human Work
Artificial intelligence is creating a less visible category of human work, from labeling road scenes to recording physical tasks for robotics. A December 2025 Oxford Economics report estimated the U.S. data annotation market at $2.7 billion to $5 billion in 2024, with potential growth to $10.3 billion to $19 billion by 2030.
- Oxford Economics estimated the U.S. data annotation market at $2.7 billion to $5 billion in 2024.
- The market could reach $10.3 billion to $19 billion by 2030.
- AI training work includes image labeling, first person video capture and robotics tasks.
- Nearly 200,000 flexible earning opportunities were directly supported in 2024.
- Large projects can require millions of verified data capture hours.
AI training moves beyond image labels
1840 & Company, a global staffing and business process outsourcing provider, says it is supporting projects that require people to produce and verify training data at scale. Its founder and CEO, Bryan DiGiorgio, said, "We can talk about the here and now. We're scaling it."
Some projects involve identifying trees, stop signs and traffic lights in images for autonomous vehicle training. Others are tied to robotics and physical tasks. Contributors may use head mounted cameras or smartphones to record themselves from a first person perspective, creating footage that AI systems can analyze frame by frame.
The recordings can show how people identify objects, move their hands, sequence actions and complete tasks. Examples described by DiGiorgio include folding clothes, making a bed, handling different objects and carrying out commercial activities. Some projects use standard phone cameras, while others require more advanced devices or sensors.
A workforce spread across countries
Much of this work is organized around project based contractors rather than permanent employees. Companies may need to coordinate specialized contributors across multiple countries for defined tasks or periods, making the activity less visible in traditional payroll data.
The U.S. Bureau of Labor Statistics excludes self employed workers from its establishment survey, which tracks nonfarm payroll employment, although its household survey captures some forms of self employment. "A good portion of this work is being done by independent contractors. There's this ballooning of jobs occurring right now that people aren't aware of," DiGiorgio said.
The Oxford Economics report estimated that the industry directly supported nearly 200,000 flexible earning opportunities in 2024. That figure illustrates the scale of work already developing around AI training, according to the source.
From finding workers to delivering verified data
1840 & Company described two ways it supports AI training programs. In a staffing arrangement, it sources, contracts and compensates contributors while the client manages the work. In a managed program, 1840 takes more responsibility for delivering an agreed amount of completed and verified data.
That can involve finding businesses where commercial activities can be recorded, distributing cameras or other equipment, coordinating contributors and handling hardware availability across locations. When devices must be shipped from another country, customs and distribution can add complexity. "The phone itself may cost what you pay a person, or even more," DiGiorgio said.
Large projects can require millions of verified data capture hours, with the work divided among several suppliers because, DiGiorgio said, "None of them can really support the volume on their own." The development of these programs points to a growing role for workforce providers that can source and manage contributors as AI systems expand into real world tasks.
PR Newswire: Human Work Behind AI Is Growing as Training Expands Into Real-World Tasks
