← All industry news

AI eyes on the line just got funded — and MRFs without them will fall behind

By The Bond4Waste editorial team·July 28, 2026·Originally reported by Resource Recycling
AI eyes on the line just got funded — and MRFs without them will fall behind
Photo by Pop & Zebra on Unsplash

Artificial intelligence just got another solid vote of confidence on the tipping floor. Greyparrot, a UK-based “waste intelligence” firm, raised $27 million to expand its Analyzer network and dataset — a move that, if you run a MRF or sell curbside processing, should accelerate your timeline for instrumenting lines and wiring that data into contracts. This isn’t about robots replacing pickers. It’s about finally having defensible, real-time composition and quality data at scale — and the leverage that gives you in pricing, staffing, and compliance.

Funding fuels an AI arms race on the line

Resource Recycling reports Greyparrot’s Series B brings total funding to $60 million, aimed at scaling its in-line camera-and-vision platform and the accompanying dataset. Translation for operators: more off-the-shelf kits, faster installs, better-trained models, and growing pressure from municipalities and brand owners to “show your data” instead of debating contamination with clipboards and one-off audits.

We’re past the pilot era. Across MRFs, near-infrared optics and QC cameras already sit over fiber, PET, OCC, and residue belts; what’s shifting is the sophistication and reliability of object-level recognition tied to SKU or brand-level insights and bale quality scoring. As vendors add devices, they sharpen models and cut false positives — which turns hourly reads into something you can defend in a council meeting or an offtake renegotiation.

Why composition data now drives contracts, EPR and bale premiums

Two forces are converging. First, contamination fees and performance guarantees in municipal contracts are getting more granular. Cities burned by 25-40% contamination don’t want quarterly audits; they want continuous data by route or day. Second, extended producer responsibility laws (think California SB 54 and programs in Oregon and Colorado) require reporting that’s tougher to satisfy with manual sampling. Real-time composition data can substantiate claims on capture rates, contamination, and outbound quality — and unlock producer funding where available.

On the commodity side, bale buyers are drawing harder lines. A facility that can prove a PET line averages under a given PVC or label contamination threshold, with time-stamped images, gets leverage for premiums or, at minimum, avoids downgrades. The same data lets you tune staffing: if you see specific contamination spikes tied to collection routes or times of day, you can redeploy QC labor to those windows instead of static staffing.

For haulers, this is coming to the invoice. If your MRF partner can attribute contamination to a set of routes or accounts using synchronized scalehouse times and line data, expect variable contamination charges — and expect them to stick when challenged. That, in turn, drives your education spend and cart-level interventions, because you’ll finally have evidence that one Tuesday route costs you an extra sorter FTE and 2% more residue.

Operational playbook: retrofits, IT plumbing, and data rights

If you’re eyeing a retrofit, the hurdles aren’t exotic but they’re real: belt coverage and lighting, camera placement and cleaning access, ruggedized enclosures, power and network runs, and the edge compute to keep classification running when the WAN blips. Don’t underestimate lens fouling; build nightly wipe-downs and weekly deep cleans into your PMs, or your beautiful dataset will drift.

On the IT side, insist on open APIs and data export, not screenshot dashboards. You’ll want to join composition reads with scale tickets, route IDs, bale IDs, and billing systems to automate contamination charges, customer scorecards, and QC staffing schedules. If the vendor can’t give you event-level data with timestamps and confidence scores, you’re buying a black box.

Lock down data ownership in the MSA. Who can resell or publish your line data? Can you move historical data if you switch vendors? What SLAs exist for model performance, uptime, and replacement parts when a camera dies mid-shift? Also decide the right coverage: a single Analyzer over residue might answer 60% of your contract questions, while full-line coverage may justify itself only if you’re chasing bale premiums or EPR reporting revenue.

Finally, consider financing. Some vendors offer subscription models that move capex to opex. That’s attractive, but read the fine print around data portability and price escalators — and make sure you’re not locked into paying for capacity you won’t use if a line is idled.

The Bond4 Tech Take

Here’s the blunt version: in the next contract cycle, MRFs without defensible, feed-and-bale data will lose price and credibility, and haulers will eat contamination fees they can’t push downstream. We’d budget 0.5%–1.0% of facility revenue annually for sensing and integration over the next three years. Start with the belts that decide money — residue and your top two fiber/plastic streams — and instrument at least 80% of throughput on those lines, or your “continuous” data will be litigated to death.

Demand API-first platforms and write data portability into the contract. If the vendor won’t commit to event-level exports, confidence scoring, and SLAs tied to fee adjustments when uptime drops, walk. Then wire the data into operations: auto-generate contamination surcharges by route, push daily contaminant alerts to driver apps, and tie bale-quality trends to offtake pricing so your sales team stops negotiating in the dark. On staffing, use the heatmaps to flex QC headcount by hour; we’ve seen that shave a sorter or two per shift without denting quality.

Expect M&A pressure. Private equity will pay a premium for “instrumented MRFs” with proven yield and quality deltas; uninstrumented plants will be valued like guessing machines. For haulers, this is your chance to renegotiate: if you can show concentrated contamination at a handful of customers, you can justify targeted education fees and avoid across-the-board price hikes.

The trap is vendor lock-in. Don’t swap human guesswork for algorithm guesswork you can’t audit. Own your data, specify performance, and make the analytics drive invoices, not just pretty dashboards.

Quick meeting

Book a meeting with Bond4Waste

Pick a time and add your details — or we'll reach out if none work.

Finding open times…
Original story
Follow Us:

Researched and drafted with AI assistance by the Bond4Waste editorial team. All credit for original reporting goes to Resource Recycling.

Related reading