AI in industrial settings has been around for years with proven machine learning (ML)-powered use cases such as defect detection, predictive maintenance, and autonomous mobile robots (AMRs). However, the applications and benefits of generative AI (GenAI) in industrial settings are not as obvious. GenAI differs from traditional AI/ML processes by using large language models (LLMs) and large visual models (LVMs) to interpret or create text, code, images, and other media. Perhaps the most promising use of GenAI at the industrial edge is in the field of human-machine interaction and specifically robotics.
Robots and Humans Are Already Working Together, Just Separately
Collaborative robots, also known as cobots, are in prominent use today. Assembly lines and fulfillment centers use a combination of cobots and human workers to assemble goods or pack and prepare shipments. Amazon recently crossed a threshold of deploying over one million robots1—many of which are AMRs or stationary robotic arms—across its fulfillment centers, alongside approximately 1.5M human workers.2
Cobots are often kept in spaces separated by cages, plexiglass, or designated zones to prevent any physical intersection between humans and machines. For example, in Amazon warehouses, human workers wear “tech vests” that allow AMRs to sense when they’re nearby and slow down, improving safety.3
In these scenarios, human workers don’t interact directly with robots; they mostly avoid each other. This is precisely where GenAI might influence the direction of a more collaborative relationship between humans and robots, by enabling situations where they occupy the same space and can engage with each other using natural language.
The Only Thing Better Than a Robot Helper Is Telling It What to Do
Verbal communication will be a linchpin in enabling shared spaces for human-robot collaboration. This is in part because humans can help assess situations and direct robots with specific tasks, but also for the simple reason that sometimes a human worker will need to tell a robot to “stop” for safety reasons.
Agility Robotics, an Oregon-based startup, has been developing a humanlike robot (a humanoid) called Digit with the express purpose of enabling safe human-robot collaboration. At the end of 2023, Agility Robotics published a demo showing Digit using an LLM to interpret and follow a written command of, “Take the box that’s the color of darth vader’s light saber and move it to the tallest tower in the front row.” The demo shows Digit’s internal monologue as it navigates pillars, boxes, and pop culture references to successfully complete the task.
Additional promotional videos show Digit responding to voice commands, although the results appear somewhat orchestrated and less improvisational. In one such video, Digit responds to the question “What’s cooking?” by retrieving a roasting pan full of GPUs from an open oven.4
Agility Robotics’ promotional materials make clear that safety isn’t guaranteed yet. Per one eBook, “The response and predictability of the outcome is not as deterministic as needed when talking about safety around people. We need to know 99.99% of the time how a robot will respond around a person, not 80% of the time.”5 While Digit robots are currently being deployed in isolated and gated spaces, the company’s stated goal is to achieve full human-robot collaboration by 2026.
But How Much Does It Cost?
Peggy Johnson, CEO of Agility Robotics, remarked in an interview that Digit’s cost was comparable to a luxury car.6 (The suggested retail price of BMW M Models range between $50,000 and $120,000.7) In another interview, she commented that an organization could expect to achieve ROI in under two years, compared to a full-time worker making $30/hr,8 which can be roughly calculated to $80,000-$90,000 per year. Given Digit’s limited feature set, the cost-to-benefit ratio probably isn’t where it needs to be to entice manufacturers, but technology costs will continue to drop as wages continue to rise.
To R2-D2 or Not to R2-D2
In a recent interview, Amazon Robotics Chief Technologist Tye Brady said, “When I think of robots, I think of R2-D2, not C-3PO. Who is in the back of the X-wing fighter when Luke Skywalker had to go into the Death Star? It was R2-D2. Why I love R2-D2—and actually, R2-D2 was the inspiration for my entire career—it’s because R2-D2 helps a Jedi be more Jedi.”9 This sentiment tracks with Amazon’s broad-scale adoption of R2-D2-like AMRs across its warehouses globally.
Perhaps unsurprisingly, Amazon has also been testing a potential application for the C-3PO-like Digit robot as well. At a RoboFab testing site in Oregon, Digit was filmed transferring bins from a shelf to a conveyor belt,10 a task more commonly delegated to human workers. This hints that Amazon might prefer to further augment or replace its Jedi workforce, in whole or in part, in the coming years.
The Future of GenAI on the Shop Floor
While the industry is still in a nascent phase of exploring GenAI possibilities, there’s a clear destination. GenAI will enhance cobots with high-level, multimodal reasoning capabilities, allowing human and robot workers to interact and help each other with tasks. The main limiters are safety and cost. In the distant future, fully automated “lights out” factories are a possibility, but the technology isn’t there yet.
What should organizations be doing now to prepare? Advanced robotics will rely on converged IT/OT networks for connectivity, control, and data analysis. As cobots and humanoids become more mainstream, manufacturers should be implementing foundational technologies now so they can readily pilot and deploy solutions, or risk falling behind.
- Bloomberg Podcasts, Amazon Marks One Million Robots Working Across Company, July 2025.
- Reuters, Amazon’s corporate workforce may shrink as AI takes over routine tasks, June 2025.
- The Week, Amazon creates ‘anti-robot vests’ to protect workers from droid collisions, Jan 2019.
- Agility Robotics, Controlling Humanoid Robots with LLMs and Voice Commands, April 2024.
- Agility Robotics, Setting The Standard, accessed July 2025.
- CNBC, Inside the world’s first humanoid factory, where robots could eventually build themselves, Oct 2024.
- BMW, BMW M Models, accessed July 2025.
- The Robot Report, Here’s what it could cost to hire a Digit humanoid, July 2024.
- Bloomberg Podcasts, Amazon Marks One Million Robots Working Across Company, July 2025.
- CNBC, Inside the world’s first humanoid factory, where robots could eventually build themselves, Oct 2024.
