Waste sorting is becoming a data problem as well as a mechanical one, with regulators and packaging companies demanding more reliable information about the materials moving through collection and recovery systems.

London based Greyparrot has raised $27 million in Series B funding to expand its computer vision platform, which uses cameras and machine learning to identify waste moving along industrial sorting lines. Technology investor Omar Mir led the round, taking the company’s reported funding beyond $50 million.

Greyparrot says its systems have recorded more than one trillion detections across facilities in over 20 countries. Customers and partners include Veolia, Biffa, FCC Environment, WM, Circular Services, and packaging businesses seeking better information about the composition and recovery of post-consumer materials.

The company reports efficiency gains of between 10% and 30% in some installations and says one site identified annual savings worth £1.5 million. Those are company and customer figures rather than independently standardised results, although they illustrate how better visibility can affect throughput, contamination, equipment settings, and the value of recovered material.

Waste facilities have operated with partial information

Traditional waste analysis depends heavily on periodic sampling. A small quantity of material is separated and inspected, producing a snapshot that may not reflect changes between shifts, suppliers, seasons, or collection routes.

Plant managers can monitor total weight and machinery performance while having less precise information about which materials are being lost, contaminating output, or passing through the wrong sorting process. Computer vision can inspect waste continuously as it moves beneath a camera, creating a larger dataset about composition and flow.

Operators can use that information to adjust equipment, compare incoming loads, identify valuable material, and investigate why output quality has changed. The economic case strengthens when commodities carry different prices or contamination causes an entire bale to be downgraded.

Identification software remains only one component of the system. Lighting, camera position, belt speed, overlapping objects, dirty packaging, and changing waste streams can affect accuracy, requiring processes for calibration, maintenance, validation, and uncertain classifications.

Regulation gives the data a second use

Greyparrot says Biffa and FCC submitted AI derived composition data that the Environment Agency accepted for quarterly statutory reporting during 2026. The development extends the value of waste analytics beyond plant optimisation by bringing continuously gathered information into regulatory processes.

Britain’s Digital Waste Tracking programme is intended to create a consistent record as material moves between producers, carriers, brokers, and treatment facilities. Better facility level data could help regulators identify discrepancies, assess recovery performance, and investigate criminal disposal or false reporting.

Using machine generated information for compliance also increases the need for common standards. Data collected by different camera systems and models must be comparable when it influences legal obligations or enforcement decisions. Regulators will need confidence in error rates, audit records, sampling methodology, model changes, and the treatment of uncertain results.

Packaging producers have a related interest because extended producer responsibility links costs to the materials placed on the market and their recyclability. Companies including Unilever, L’Oréal, and Kenvue use Greyparrot’s Deepnest platform to analyse material outcomes, according to the company.

Such data can influence packaging design, although it cannot resolve fragmented incentives across production, collection, sorting, and reprocessing. A package may be technically recyclable while failing to reach the correct plant, or it may be accepted in one local system and rejected in another.

Greyparrot intends to use the funding to expand internationally and develop further products, with a stated aim of helping to abate more than one million tonnes of waste by 2030. Progress will depend on measured operational change rather than the number of objects detected.

The larger opportunity lies in turning waste facilities into measurable production environments. Once material data influences equipment settings, packaging design, statutory reports, and producer charges, the cameras above a conveyor become part of industrial and regulatory infrastructure rather than an optional analytics layer.