The truck still doesn’t know what’s in the dumpster, but RTS does.
A waste truck rolls up to a container on Tuesday because the calendar says so, not because the container sits full. Sometimes the container sits half-empty. Sometimes the container overflows and stays that way for two days, and nobody found out until a tenant complained. Either way, the truck runs the route blind, and until recently, so did everyone downstream of it.
That approach is the model most of the waste industry still runs on: fixed schedules, static routes, and a driver’s memory standing in for real data. The approach has worked for decades because there wasn’t a better option. A better option exists now.
Artificial intelligence, in the specific and unglamorous sense of sensors, computer vision, and route algorithms tied together into one system, already changes how collection planning and verification happen. I’ve spent twenty-five years around this industry, long enough to put it plainly: the shift isn’t coming. The shift arrived, and it compounds faster than most operators outside the tech-forward corner of this business realize.
RTS Turned Trash Into a Data Feed
RTS is the sharpest example I’ve watched of what full build-out looks like, not a bolt-on fix. Pello, the sensor line RTS brought fully in-house through the RecycleSmart acquisition, sits inside containers and measures fill level with ultrasonic detection; combine that with computer vision at the point of pickup, and a hauler suddenly knows two things nobody reliably knew before: how full a container actually is, and whether the truck actually serviced that container.
That verification detail matters more than expected. Missed pickups and disputed service calls long acted as a quiet tax on hauler-client relationships; the driver insists he serviced the container, the client insists otherwise, and somebody eats the cost of the argument. Pickup verification removes the argument. The truck either lifted the container or didn’t, and the sensor and camera both agree on which.
That 2023 RecycleSmart acquisition is where Pello came from, and RTS now runs machine-learning monitoring across a field of thousands of rugged sensors that have to survive compaction, weather, and general abuse inside a dumpster. Keeping that many devices online, calibrated, and reporting accurately amounts to a distinct engineering problem, one most people never consider because the task looks invisible when it works. When the system works, the output feeds into a centralized platform that handles dispatch, billing, and dispute resolution from data instead of from a phone call. The platform is the missing link the industry lacked for a very long time.
The Hauler Is Still the One Who Shows Up
Abstract talk about AI and waste tends to skip past one fact: none of this replaces the hauler. The trucks still roll, the drivers still lift, the routes still have to get run in the rain. The AI shift changes the planning, the verification, and the paperwork that used to eat hours a driver could spend on another stop.
Driver shortages persist, turnover costs money, and route fatigue burns out good people faster than most fleets replace them. [Algorithmic routing balances workloads across a fleet instead of leaving route assignment to habit or seniority, doing the boring but valuable work of cutting mileage nobody needed to drive in the first place. Fewer wasted miles means less fuel burned and less wear on the trucks; a shift also ends closer to on time.
The proof-of-service piece does something else worth naming: driver protection. A contested pickup used to come down to whoever’s account sounded more credible. A timestamp and an image settle the question now. On-truck AI vision gives haulers “real-time visibility into bin contents, access issues, and service verification” that drivers and dispatchers simply lacked before; unfamiliar stops stop being a guessing game, and a route running long stops being a mystery.
Then there’s the back office, where a lot of hauler resourcing quietly disappears. Billing reconciliation, contract auditing, compliance paperwork: none of these tasks collects a single can, and every one of them needs a person. RTS’s waste broker model offloads that administrative load so hauling partners can put people and capital toward trucks and routes instead of spreadsheets. That administrative load is the resourcing story in one sentence: less time spent proving the work happened, more time spent doing it.
What Clients Get Back
A property manager or restaurant group feels this from the other direction. A property manager or a restaurant group cares little about ultrasonic sensors as a technology; these clients care about not paying for a pickup that never needed to happen.
Fill-level data lets a hauler rightsize a container schedule instead of guessing at it — swapping a fixed weekly service for one tied to how fast the container actually fills. Contamination makes up the other half of the cost problem; it stays invisible until a load fails inspection at the recycling facility and someone eats a penalty fee. Optical recognition cameras mounted on collection vehicles catch contamination at the point of pickup, which is the only point where catching it actually helps.
And then there’s reporting. Enterprise clients with multi-site portfolios need diversion and ESG data they can hand to a sustainability team or a regulator without assembling it by hand from twelve different hauler invoices. A connected platform that already has the sensor and pickup data can generate that reporting automatically, because the data existed the moment the truck did its job. Container size matters. Fill rate matters. Material type matters. Service confirmation matters. Every one of those data points used to require a phone call; now each is a byproduct of the collection itself.
None of This Works Without Clean Data
Honesty here means naming where the friction actually lives: not in the algorithm. The friction sits in the hardware and the data pipeline underneath it. A sensor that takes in garbage data produces garbage output, and a device sitting inside a dumpster gets crushed, rained on, and occasionally stolen. Building a sensor network that survives that environment and still reports clean, consistent numbers is unglamorous work, and that work determines whether any of the rest of this holds up.
The other piece is human, not technical: this only functions with a driver still in the loop. AI flags a likely contamination event or a container trending toward overflow; a person decides what to do about the situation. Route optimization suggests the sequence; a driver still knows that the alley behind the loading dock floods every time it rains and adjusts accordingly. Software doesn’t have that memory. Software supports the driver’s judgment; it doesn’t replace it. Selling this as a replacement for driver expertise misreads how a truck actually moves through a city.
The Money Is Already Moving
RTS raised $40 million led by Edison Partners in 2025 to keep building this out (the same reporting also covers litigation tied to the deal, worth knowing before treating the raise as a clean signal), and the funding chases a real thesis: waste collection generates enormous amounts of usable data, and almost nobody captured it until now. Trash turns out to be a supply chain hiding in plain sight, and somebody finally started measuring the flow.
For a hauler running on tight margins and a tighter labor pool, that’s not an abstract industry trend. The upside is fewer disputed pickups, fewer wasted miles, less time buried in paperwork, and a business that grows without a proportional hire for every new account. For clients, the payoff is a bill that actually reflects what happened at the container. Different problems, same fix; this kind of alignment rarely comes along in this business.
The trucks still have to show up. What changes is how much the drivers — and everyone paying for the ride — finally get to know.
