The Jobs No One Should Have to Do
Every year, roughly 5,000 workers die on the job in the United States. Tens of thousands more die from occupational diseases. Globally, the International Labour Organization estimates 2.3 million work-related deaths annually—6,300 per day.¹
The most dangerous jobs are often essential: mining the materials that power the economy, building and maintaining infrastructure, responding to disasters, working in extreme environments. These jobs are also often undesirable for other reasons—physically demanding, unpleasant, repetitive, or remote from family and community.
For decades, the promise of robotics has been to take over these jobs—to send machines instead of humans into mines, onto construction sites, into burning buildings, and up electrical towers in storms. That promise has been partially fulfilled: robots handle tasks in factories that once maimed workers; robots explore deep ocean trenches and nuclear reactors. But the frontier has been stubborn. Unstructured environments, varied tasks, and the need for human judgment have kept most dangerous work in human hands.
That's beginning to change. AI enables robots to perceive and navigate complex environments. Better actuators and batteries extend robot capability. Declining costs make automation economically viable in more contexts. The next decade may see significant expansion of robot capability into the dangerous, dirty, dull, and distant work that humans still do today.
This chapter explores that expansion: what robots can do now, what's emerging, and what it means when machines can go where humans shouldn't.
2026 Snapshot — Current Deployments
Industrial Robotics
Factory robots remain the largest robot market:
- 3.5+ million industrial robots operating globally
- Automotive, electronics, metal/machinery are largest sectors
- China is largest market and manufacturer
- Collaborative robots ("cobots") work alongside humans²
Common applications:
- Welding (arc, spot, laser)
- Assembly (component insertion, fastening)
- Material handling (palletizing, loading/unloading)
- Painting and coating
- Inspection (visual, dimensional)
Limitations: Industrial robots work in controlled environments. They're programmed for specific tasks in predictable settings. The "4D" work that happens in unstructured environments is different.
Construction Robotics
Current state: Limited deployment; most construction remains manual.
Active applications:
- Bricklaying robots (Hadrian X, SAM100): Laying bricks faster than human masons; deployed at small scale
- 3D printing: Large-scale concrete printing for structures (ICON, Apis Cor); demonstration projects and limited commercial use
- Demolition robots: Remote-controlled machines for hazardous demolition
- Drilling and excavation: Automated systems for tunneling and foundation work
- Drones: Surveying, inspection, progress monitoring
Challenges: Construction sites are unstructured, changing, and require coordination across trades. Most work still requires human dexterity and judgment.
Mining Robotics
Current state: Significant automation in large-scale mining; humans still do most underground work.
Active applications:
- Autonomous haul trucks (Caterpillar, Komatsu): Operating at many large mines; thousands deployed
- Autonomous drilling: Blast hole drilling in surface mining
- Remote operation: Controlling equipment from distant operations centers
- Inspection robots: Monitoring infrastructure, detecting hazards
Challenges: Underground mining involves cramped, variable environments. Ventilation, dust, and communications are difficult. Human miners still do most extraction work.
Inspection and Maintenance
Current state: Growing deployment for dangerous inspection tasks.
Active applications:
- Drones for power line, bridge, building, and tower inspection
- Crawling robots for pipeline, tank, and vessel inspection
- Underwater ROVs for subsea infrastructure
- Climbing robots for wind turbine blade inspection
Economic driver: Inspection previously requiring scaffolding, confined space entry, or working at height can be done safely by robots.
Disaster Response
Current state: Limited but growing; most disaster response is still human.
Active applications:
- Bomb disposal robots: Widely used by military and police
- Search and rescue drones: Finding survivors in rubble, wilderness
- Fire reconnaissance drones: Mapping wildfires, providing situational awareness
- Nuclear inspection robots: Working in radioactive environments
Challenges: Disaster environments are the most unpredictable—collapsed structures, fire, water, debris. Robots lack human adaptability. Reliability is critical when lives depend on it.
Space and Extreme Environments
Space robotics (see Section D):
- Rovers on Mars (Perseverance, Curiosity)
- Robotic arms on ISS
- Satellite servicing (MEV)
Underwater:
- ROVs for deep-sea work (oil/gas, research, salvage)
- Autonomous underwater vehicles (AUVs) for survey and inspection
Polar and high altitude: Monitoring stations, scientific robots
Notable Players
Construction
Built Robotics: Autonomous excavators and other heavy equipment for earthmoving.
Dusty Robotics: Layout robots marking positions for construction crews.
Canvas: Drywall finishing robots.
ICON: 3D printing of homes and structures.
Boston Dynamics (Spot): Quadruped robot for construction site inspection.
Mining
Caterpillar: Autonomous haul trucks (over 600 operating); broad mining automation portfolio.
Komatsu: Autonomous haulage system (AHS); underground automation.
Sandvik: Underground drilling and loading automation.
Epiroc: Mining equipment and automation systems.
Rio Tinto: Mining company with extensive automation deployment (Mine of the Future program).
Inspection
Flyability: Drones for confined space inspection.
Gecko Robotics: Wall-climbing robots for tank and infrastructure inspection.
SkySpecs: Drone inspection for wind turbines.
SeaRobotics: Marine inspection and survey vehicles.
Sarcos Robotics: Exoskeletons and inspection robots.
Disaster Response
Boston Dynamics: Spot deployed for hazmat inspection, fire departments.
Clearpath Robotics: Unmanned ground vehicles for research and industrial applications.
DARPA programs: Funding development of disaster response robots (e.g., DARPA Robotics Challenge).
General Industrial
Fanuc, ABB, KUKA, Yaskawa: Industrial robot leaders.
Universal Robots: Collaborative robots.
Realtime Robotics: Motion planning for industrial applications.
Technology Advances Enabling Expansion
Perception
Computer vision improvements enable robots to work in unstructured environments:
- Object recognition in varied conditions
- Depth perception and 3D mapping
- Semantic understanding of scenes
- Real-time processing at low power
LiDAR and radar costs have fallen while capability has improved.
Fusion of multiple sensor types provides robust perception.
Navigation and Mobility
Autonomous navigation in complex environments:
- SLAM (Simultaneous Localization and Mapping)
- Path planning around obstacles
- Dynamic environment adaptation
Locomotion improvements:
- Legged robots (Boston Dynamics, others) handling rough terrain
- Better wheeled and tracked designs
- Improved climbing and maneuvering
Manipulation
Grasping in varied conditions:
- Soft grippers that handle fragile or irregular objects
- Multi-finger dexterous hands (still limited but improving)
- Force sensing and control
General manipulation remains challenging but advancing.
AI Integration
Foundation models enabling new capabilities:
- Understanding natural language instructions
- Generalizing from demonstrations
- Adapting to new tasks without explicit programming
Imitation learning: Robots learning from human demonstration rather than programming.
Simulation: Training robots in virtual environments before physical deployment.
Power and Endurance
Battery technology (see Chapter 10) enables longer operation:
- Higher energy density
- Faster charging
- Longer cycle life
Tethered power for stationary applications provides unlimited endurance.
Communications
5G and beyond enables remote operation with low latency.
Mesh networking for operation in infrastructure-poor environments.
Edge computing reduces bandwidth needs for autonomous operation.
The 4D Categories
Dangerous Work
What qualifies: Work with significant risk of death, injury, or occupational disease.
Examples being automated:
- Mining (cave-ins, equipment accidents, lung disease)
- Construction (falls, struck-by incidents, electrocution)
- First response (fire, hazmat, search and rescue)
- Electrical work (electrocution, falls)
- Logging (struck by trees, equipment accidents)
The promise: Every job a robot does is a worker who doesn't risk their life.
The reality: Most dangerous jobs still require human judgment and dexterity. Robots can take over specific tasks; full replacement is further away.
Dirty Work
What qualifies: Work that is unpleasant, unsanitary, or exposes workers to hazardous materials.
Examples being automated:
- Waste handling and recycling
- Sewer and tank cleaning
- Industrial cleaning in hazardous environments
- Animal processing
- Decontamination and remediation
The promise: No one should have to work in sewage or toxic environments if machines can do it.
The reality: Many dirty jobs also require adaptability and judgment. Automation is advancing but incomplete.
Dull Work
What qualifies: Highly repetitive work that provides little engagement or growth.
Examples being automated (covered in Chapter 15):
- Warehouse picking and packing
- Sorting and categorization
- Security patrol and monitoring
- Data entry (digital, not physical)
The overlap: Much dull work is also dangerous (repetitive stress injuries, heat exposure, chemical exposure in some settings).
Distant Work
What qualifies: Work in locations remote from workers' homes and communities.
Examples being automated:
- Offshore oil and gas operations
- Remote mining sites
- Agricultural work in distant locations
- Infrastructure in remote areas
The promise: Workers could operate robots from home rather than spending months at remote sites.
The reality: Remote operation is increasingly feasible. Full automation is harder.
The Path Forward
Near-Term Likely (2026-2032)
Construction automation expands: More bricklaying, 3D printing, and earthmoving robots deploy. Drones and inspection robots become standard. Humans still do most skilled trades work.
Mining automation deepens: More autonomous haul trucks; drilling automation expands; remote operation centers become standard. Underground mining begins to see more automation.
Inspection robots proliferate: Most infrastructure inspection that previously required scaffolding or confined space entry shifts to robots. Significant cost and safety improvements.
Disaster response capability grows: Robots routinely deployed for hazmat, fire reconnaissance, and initial search. Humans still lead complex rescue operations.
Plausible (2032-2040)
Humanoid robots in construction: General-purpose robots begin handling varied construction tasks. Not replacing all workers, but handling more operations.
Underground mining transformation: Significant portions of underground mining become autonomous or remote-operated. Fewer humans underground.
Maintenance automation: Robots that can identify problems and make repairs, not just inspect. Predictive maintenance combined with robotic intervention.
Offshore and remote operations: Minimal human presence at remote installations. Operation centers in population centers control remote robots.
Wild Trajectory (2040+)
Most dangerous work eliminated: Robots handle the vast majority of tasks that currently cause workplace fatalities. Human work-related deaths drop dramatically.
Remote resource extraction: Mining, oil/gas, fishing, and other resource industries operate with minimal on-site human presence.
Disaster response transformation: Robot teams handle most emergency response. Humans manage from safety.
Construction is largely automated: From site preparation through finishing, robots handle most physical construction work.
Second-Order Effects
Worker Safety
The direct benefit: Fewer injuries and deaths if robots do dangerous work.
Current numbers: ~5,000 US workplace deaths/year; ~2.8 million injuries. Globally, far higher.³
Potential impact: If robots handle the most dangerous 20% of jobs, workplace fatalities could fall significantly.
The transition period: Workers who currently do dangerous jobs need transition to new roles. The benefit of robot safety only matters if displaced workers find safe alternatives.
Labor and Economics
Jobs affected: Mining, construction, and similar fields employ millions. Not all jobs would be eliminated, but many would change.
Skill shift: Operating and maintaining robots requires different skills than manual work. Retraining is necessary.
Geographic shift: Remote operation enables workers to live anywhere. Remote communities that depend on on-site work could be disrupted.
Wage effects: If robots can do dangerous work, the wage premium for dangerous work may disappear. Workers would need to find value elsewhere.
Environmental Operations
Hazardous cleanup: Robots can work in environments too dangerous for humans—radioactive sites, toxic waste, contaminated facilities.
Resource extraction: Remote and automated extraction could access resources currently too dangerous or remote. Environmental implications vary.
Disaster remediation: Robot capability to work in disaster environments could accelerate recovery.
Accessibility
Remote work for physical jobs: Workers with disabilities that prevent being on-site could operate robots remotely. Physical labor becomes accessible to more people.
Aging workforce: Remote operation could extend working life for physical occupations.
Risks and Guardrails
Safety in Transition
Robot reliability: Robots that fail in dangerous environments can cause harm—to other workers, to property, or by failing to perform critical functions.
Shared environments: During transition, humans and robots will work together. Safety systems must prevent robot-human incidents.
Guardrails: Extensive testing; redundant safety systems; clear protocols for robot failure; gradual deployment in mixed environments.
Liability and Accountability
When robots cause harm: Who is responsible when a robot injures someone or damages property? Manufacturer, operator, programmer?
Insurance frameworks: Existing workers' compensation and liability frameworks don't clearly cover robot operations.
Guardrails: Clear liability frameworks; insurance requirements; incident reporting and investigation standards.
Job Transition
Displacement risk: Workers in dangerous occupations may lack skills for new roles. Geographic displacement if remote operation centralizes.
Wage pressure: If robots can do dangerous work, workers lose bargaining power.
Guardrails: Transition support and retraining; regional economic support for affected communities; policies that share productivity gains.
Security
Critical infrastructure: Robots operating infrastructure could be targets for cyberattack or physical disruption.
Dual use: Technologies for dangerous work could be adapted for harmful purposes.
Guardrails: Cybersecurity requirements; access controls; monitoring for misuse.
The AI Acceleration Factor
AI is the key enabler for expanding robot capability in unstructured environments:
Perception: AI enables robots to understand complex environments—not just detect obstacles but understand what they're seeing.
Planning: AI enables robots to figure out how to accomplish tasks in varied conditions, not just follow scripts.
Learning: AI enables robots to improve from experience and adapt to new situations.
Language: Foundation models could enable instructing robots in natural language, reducing the need for specialized programming.
Simulation: AI-generated simulations provide training environments for robots to learn before physical deployment.
The trajectory: As AI improves, robots become capable in more unstructured environments. The timeline for automating dangerous work accelerates.
Conclusion
The case for automating dangerous work is simple: no one should die or be injured doing work a machine could do. The technology to achieve this is advancing, driven by AI, better sensors, and improved robotics hardware.
The path is not straightforward. Unstructured environments remain challenging. Economic incentives don't always favor safety. Transition for affected workers requires attention and investment.
But the direction is clear. The most dangerous jobs—mining, construction, disaster response, maintenance at height, work with hazardous materials—will increasingly be done by machines. The question is how fast, how completely, and whether society manages the transition to ensure that workers displaced from dangerous jobs find good alternatives.
The goal is not to eliminate work but to eliminate unnecessary danger. If achieved, the next generation may find it strange that earlier generations ever sent humans into situations where machines could have gone instead.
Endnotes — Chapter 24
- Bureau of Labor Statistics reports approximately 5,000 US workplace fatalities annually. ILO estimates 2.3 million global work-related deaths per year including occupational disease.
- International Federation of Robotics tracks industrial robot installations. Global operational stock exceeded 3.5 million units as of 2023; China leads in both production and installation.
- BLS data shows approximately 2.8 million nonfatal workplace injuries and illnesses annually in the US private sector.
- Caterpillar autonomous haul truck fleet exceeded 600 units as of 2024; billions of tonnes moved without human drivers.
- ICON and other construction 3D printing companies have demonstrated homes and structures printed in days rather than months; commercial deployment is limited but growing.
- Boston Dynamics' Spot robot deployed at numerous construction sites, fire departments, and industrial facilities for inspection applications.
- Built Robotics autonomous excavators operating at construction sites across North America; retrofit kits convert existing equipment to autonomous operation.
- DARPA Robotics Challenge (2015) tested disaster response robots in tasks including door opening, valve turning, stair climbing, and vehicle operation. Results showed significant capability gaps.
- Rio Tinto's Mine of the Future program pioneered large-scale autonomous trucking; operations in Pilbara region of Australia run significant portions autonomously.
- Remote operation centers enable controlling mining equipment from hundreds or thousands of kilometers away; workers can operate heavy equipment from urban offices.