The ideal of sustainability is to embed “green” attributes at the product lifecycle’s starting point — the design phase — and to build a closed-loop supply chain that enables resource circulation and regeneration. Yet designers often lack comprehensive knowledge of materials’ environmental properties, and supply-chain planners struggle to simulate complex, dynamic circular-economy networks.

This perspective proposes a human–machine collaborative innovation environment powered by an LLM. In product design, a designer can make natural-language requests to an LLM assistant (for example, “find a biodegradable alternative material with comparable strength for this phone casing”). The LLM queries built-in green-materials databases and academic literature, recommends alternatives (e.g., high-performance bio-based plastics), and compares their mechanical properties, cost, carbon footprint, and recyclability to inspire design choices and reduce environmental impact at the source.

At the supply-chain level, the LLM and intelligent optimization algorithms jointly construct a digital-twin simulator. The LLM consolidates and interprets textual information about macroeconomic policy, market trends, and consumer recycling behavior to forecast external changes. Under rules and constraints defined by the LLM, IO algorithms dynamically optimize the complex closed-loop supply network — including forward manufacturing and distribution and reverse processes such as collection, remanufacturing, and resale — to identify operating strategies that maximize resource utilization, minimize total cost, and reduce carbon emissions. This enables enterprises to confidently respond to risks such as “green tariffs” and material shortages, and to advance toward a true circular economy.