It will pick all day, 24 hours a day, seven days a week.
In the dust-laden sorting halls of east London and beyond, an old tension between human endurance and industrial demand is reaching a turning point. Waste management companies, long struggling to retain workers in conditions that injure and exhaust at rates far above other industries, are now training humanoid robots to stand at the conveyor belt in their place. The machines do not tire, do not grieve, and do not quit — and in that indifference lies both the promise of safer workplaces and the quiet displacement of thousands of people whose labor has long gone unnoticed.
- Recycling plants are caught in a staffing crisis so severe that four in ten workers leave every year, driven out by dust, noise, injury, and the relentless pace of the belt.
- A humanoid robot named Alpha is being trained at a Rainham plant using VR-captured human movement and millions of daily data points — not yet ready, but learning fast.
- Companies like AMP and Glacier are pursuing parallel automation paths, with robotic systems already sorting at eight to ten times the speed of human workers.
- Industry researchers and plant operators alike now agree that full automation is not a question of if, but when — and the economic logic is difficult to argue against.
- The unresolved tension is human: promises of retraining into maintenance roles sound reasonable at one plant, but may not hold when automation scales across an entire industry.
At a recycling plant in Rainham, east London, Sharp Group processes 280,000 tonnes of mixed waste each year under conditions that are loud, dusty, and genuinely dangerous. With a 40% annual staff turnover and injury rates running 45% above other industries, the company has begun training a humanoid robot named Alpha to take over sorting duties on the conveyor line.
Alpha is being developed by TeknTrash Robotics, a British firm working with technology from China's RealMan Robotics. The humanoid form is a deliberate choice — by mimicking human proportions and movement, the robot can slot into existing plant infrastructure without costly redesigns. Training is slow and data-intensive: workers wearing VR headsets demonstrate correct picking techniques, generating millions of data points daily through a system called HoloLab. TeknTrash founder Al Costa is candid that the technology requires extensive preparation before it becomes genuinely useful.
For finance director Chelsea Sharp, the appeal is straightforward. A robot works continuously, without fatigue, sick leave, or safety incidents. Other companies are pursuing similar ends through different means — AMP uses air jets and AI to sort at up to ten times human speed, while California startup Glacier has built robotic arms capable of learning from over a billion items, designed to work in smaller facilities as well as large urban plants.
Academic voices, including Yale's Marian Chertow, agree that robotics paired with AI vision systems represent the most promising path forward for material recovery, worker safety, and economic viability in recycling. The direction of travel is clear.
What remains unresolved is the human cost of that journey. Sharp and others promise that workers will be retrained to maintain and oversee the new systems, moving away from hazardous conditions. But as automation scales across the industry, the arithmetic grows harder to ignore — there may simply not be enough oversight roles to absorb the thousands of agency workers whose hands have always done this work.
At a recycling plant in Rainham, east London, the air hangs thick with dust and the conveyor belts never stop moving. Sharp Group, a family-run waste management company, processes 280,000 tonnes of mixed recycling annually here—everything from old shoes to concrete blocks to VHS tapes. The work is relentless, the conditions are harsh, and the company struggles to keep people on the job. With a 40% annual staff turnover and injury rates running 45% higher than other industries, Sharp Group is now turning to an unexpected solution: a humanoid robot named Alpha.
The waste-sorting industry has always been difficult to staff. The dust is pervasive, the noise from hoppers and conveyor belts is constant, and the work itself is genuinely dangerous. Work-related injuries and illnesses occur at rates well above the national average, and fatalities are a sobering multiple of what workers face in other sectors. Line supervisor Ken Dordoy describes the challenge plainly: pickers simply cannot sustain the pace. The company rotates workers through different materials every 20 minutes and periodically stops the belt to give people a break, but even these measures cannot stem the exodus.
Alpha is being trained by TeknTrash Robotics, a British firm adapting technology built by RealMan Robotics in China. What makes this approach different from earlier automation efforts is the humanoid form itself. TeknTrash founder Al Costa argues that by mimicking human movement and proportions, the robot can integrate into existing plants without requiring expensive redesigns to machinery and infrastructure. When I visited, Alpha was still in training—guided through arm movements by technicians, learning to identify items on the conveyor and lift them correctly. A worker wearing a VR headset demonstrated successful picking techniques, generating millions of data points daily through a system called HoloLab. Multiple cameras track every movement, every failure, every unpicked item that passes by. The training is painstaking, but Costa insists this is how the technology actually works in practice. "The market thinks these robots are ready-made," he explained. "But they need extensive data to be effectively useful."
Chelsea Sharp, the plant's finance director and granddaughter of the company founder, sees the appeal clearly. A humanoid robot can work continuously—24 hours a day, seven days a week, with no holidays, no sick days, no fatigue. It stays in place and performs the same task indefinitely. The alternative would be building entirely new facilities or retrofitting existing ones with specialized equipment, a far costlier proposition. Other companies are pursuing different automation strategies. AMP, based in Colorado, uses air jets to guide items into sorting chutes, paired with AI systems that improve sorting accuracy and speed—their robots operate at eight to ten times the pace of human workers. Glacier, a California startup founded by Rebecca Hu-Thrams, uses mounted robotic arms and AI vision to handle the enormous variability of trash, learning from more than a billion items to improve performance over time. Hu-Thrams notes that her company has designed systems to work not just in large urban plants but in smaller, semi-rural facilities operating on tighter budgets.
All three companies agree on one point: the human-intensive model is no longer sustainable. Academic researchers studying waste processing concur. Marian Chertow of Yale University argues that robotics combined with AI-driven vision systems offer the greatest potential for improving material recovery, worker safety, and economic competitiveness in recycling. The shift to automation appears inevitable, even necessary.
But the human question lingers. Sharp acknowledges that the work is genuinely unappealing—dirty, loud, dusty, physically demanding. Robots are indifferent to these conditions. The company's answer is that workers will be upskilled to maintain and oversee the robots, moving them away from hazardous environments and heavy lifting. Yet as automation scales across the industry, it remains unclear whether there will be enough maintenance and oversight roles to absorb the thousands of agency workers currently sorting waste on conveyor belts. The promise of retraining is real, but so is the possibility that many workers will simply be displaced, their labor no longer needed in a facility where robots work without complaint, without rest, and without the human costs that have always defined this work.
Citas Notables
The market thinks these robots are ready-made, that all you need to do is plug them in and they will work flawlessly. But they need extensive data to be effectively useful.— Al Costa, TeknTrash Robotics founder and CEO
Robotics coupled with AI-driven vision systems offers the greatest potential for improving material recovery, worker experience, and economic competitiveness in the recycling sector.— Marian Chertow, Yale University
La Conversación del Hearth Otra perspectiva de la historia
Why humanoid robots specifically? Why not just use more specialized machines like the air-jet systems?
Because a humanoid can fit into a plant that was built for humans. You don't have to tear everything apart and rebuild. You just plug it in and it learns the job.
But the training takes forever, doesn't it? Thousands of data points, VR headsets, all of that.
Yes, but once it's trained, it never gets tired, never calls in sick, never asks for a raise. It works 24/7. For a company bleeding money on turnover, that's compelling.
The company says workers will be retrained for maintenance roles. Do you believe that?
I think they mean it. But there are 24 agency workers on that line right now. How many maintenance jobs will actually exist? That's the question nobody's answering yet.
Is this inevitable?
The academics think so. The companies certainly do. The conditions are so bad and the turnover so high that something has to give. Whether it's robots or better wages and working conditions—that's the real choice, and it seems like the industry has already decided.