TOMRA increases its investment in PolyPerception to a 51% majority stake, further integrating real-time waste stream analytics with automated sorting operations.

The new AI-agent platform moves beyond traditional reporting tools by allowing operators to interact with plant data through natural language queries. Users can ask operational questions directly and receive immediate answers supported by real-time data analysis and performance insights. The platform can also generate custom quality reports, create operational alerts and support decision-making without requiring operators to work through complex datasets manually.

Additional features include:

• Similarity search for identifying visually similar objects such as batteries or electronic vapes

• Text and brand search for tracking specific materials and packaging types in real time

• Open integration with existing plant management and reporting systems

TOMRA also expanded its GAINnext™ ecosystem with three new deep learning applications targeting complex sorting challenges.

• The new PET tray application achieves purity levels above 95%, helping recyclers separate food-grade trays from other PET packaging streams.

• In metals recycling, TOMRA introduced a high-precision application for detecting copper-steel composites such as motor armatures, supporting higher-value material recovery.

The company also launched a high-throughput used beverage can (UBC) aluminum sorting application for the European market, delivering up to 98% purity and significantly higher throughput compared to manual sorting processes.

According to TOMRA, the latest developments mark a broader shift toward intelligent, connected recycling plants where AI links operational data directly with physical sorting actions.