Inside the Circle is a monthly column on circular economy trends transforming source reduction and recycling.
People fear artificial intelligence in part because they see it as a threat to jobs. If a camera and an algorithm can sort a PET bottle faster than a person, the reasoning goes, then the person on the line is the cost the technology exists to remove.
That is one way to use AI. But it is not the most valuable one.
Economists who study automation — Daron Acemoglu, David Autor and Simon Johnson among them — draw a line between two kinds of technology. One removes the human being from the process. The other removes the drudgery, the guesswork and the bottlenecks from the human being, and hands the capability back. They call the second kind pro-worker AI, and it is the version our industry should be building toward. Because in recycling, the bottleneck has almost never been the worker. It has been contamination, confusion and materials we could not economically sort, no matter our skill or speed.
Start with contamination — the quiet tax that degrades nearly every load. Intuitive AI’s Oscar, an assistant that retrofits to almost any bin, watches the moment of disposal and tells a person, in real time, which stream an item belongs in. It does not sort the waste for them; it makes the upstream human decision correct more often, before the mistake is baked into the bale — a very different intervention from adding another sorter downstream.
Just as important, it measures what is actually being thrown away, item by item, at the point the material enters the stream. You can only manage what you can measure, and for most of this industry’s history, waste at the bin has been invisible. Oscar is already running where contamination is worst and the stakes are public — stadiums, airports, campuses, event venues — turning the messiest moment in the whole system, an untrained hand hovering over three identical lids, into a correct decision and clean data.
SuperCircle works the other end of the pipe, on a stream much of the industry has written off: textiles. Apparel and footwear are built from blended synthetics — polyester, nylon, elastane — that conventional systems cannot economically separate, so most of it is landfilled, incinerated or shipped abroad. SuperCircle’s AI-driven sortation reads incoming garments and routes each to its best next life, whether fiber-to-fiber recycling, component recovery, or reuse, across dozens of downstream pathways. It also runs the reverse logistics that let brands like Guess, J.Crew and Reformation take product back at all. A $24 million Series A late last year is a signal worth reading: recovering the “unrecyclable” is starting to look like a business rather than a pilot.
Notice what these tools have in common. Neither replaces a recycling professional. Each removes a barrier that has capped what the professional and the system could accomplish — poor inbound quality, confusion at the bin, materials with no viable path forward. That is the pattern worth adopting. The AI worth having in this industry is not the kind that promises to run the plant with fewer people. It is the kind that lets the people and plants we already have capture more material, at higher quality, from streams that used to escape us entirely.
This matters beyond efficiency, and it matters for the reason the circular economy exists at all. A circular economy is not simply a linear one run in reverse. It depends on material actually finding its way back and moving forward to its highest use — cleanly, affordably, at scale. Every point where a person guesses wrong, or a material has nowhere to go, is a place where the circle breaks and the loop reverts to a line. Tools that repair those breaks are doing circular work, whether or not they put the word on their website.
The caution is real, and worth stating plainly. AI can just as easily be pointed the other way: used to squeeze more output from the same people, to concentrate control of hard-won waste data in a few hands, or to sell the appearance of diversion without its substance. The technology does not choose its purpose. We do — in what we buy, what we reward, and what we ask it to optimize. A smarter linear economy is still a linear economy. The goal is not a more efficient machine. It is a system that behaves less like a machine and more like a living cycle.
That choice is the conversation this industry should be having now, while the tools are still young enough to shape — sorting where AI is genuinely advancing circularity on the ground from where it is only industrialism with better cameras.
The recycling community has spent a decade being told that AI will change everything. The more useful question is the one worth pressing every vendor, every pilot, and every proud new deployment to answer — how will we use AI, to change things for good?
