At the Lenovo Smart Hub Salon, Lenovo recently broke down the latest version of its Lexiang Enterprise Super Agent, which has been upgraded to version 4.0. The company also disclosed fresh operational data: the platform's monthly active users have surpassed 7 million, with over 1 million daily interactive requests processed, and cumulative sales driven have exceeded 5 billion yuan.
AI is reshaping the business landscape at an unprecedented pace, evolving from a supplementary efficiency tool into a core strategic engine for driving business growth. As AI agents move beyond theoretical concepts and enter the phase of large-scale practical deployment, companies face a critical challenge: in real-world business environments characterized by long chains, multiple systems, and high uncertainty, how can AI be deeply embedded to perform effectively, transforming cutting-edge technology into a tangible, scalable, and controllable new productive force?
According to Sun Haining, Director of Human Resources Strategic Design at Lenovo China, the company is systematically recruiting "AI natives." Over the past year, AI expert positions have accounted for nearly 40% of all hires in Lenovo China's marketing department. Almost all job interviews now assess candidates' AI thinking and "AI capabilities." Lenovo has also built a comprehensive training system for its employees. These deep organizational and talent transformations have yielded quantifiable business results: in the last fiscal year, the Lexiang agent, in collaboration with 46 vertical-scenario agents forming an "agent army," helped Lenovo China's direct-to-customer revenue grow by 33% year-on-year, accounting for over 80% of total revenue. This has truly made it a "silicon-based combat force" driving business growth.
In global rankings, Lenovo has achieved several leading results: first place in the GAIA benchmark (with an overall score of 91.36%) and second place in the Terminal Bench 2.0 (with an accuracy rate of 84.5%). "We have packed all these globally leading, self-developed technologies into the enterprise-level Agent Harness. This creates a unified operational boundary, ensuring the agent has the ability to reflect and correct its actions during long-chain execution," shared He Ming, R&D Director of Agentic AI at the Lenovo Research Institute AI Lab.
On the front of autonomous evolution, the Lexiang agent relies on two core mechanisms: memory retrieval and Skill evolution. The QueryLink technology improves semantic similarity by 15.4% and reasoning accuracy by 7%, while the Lemon Agent technology boosts task success rates by 30% and generalization ability by 20%. In terms of multi-agent collaboration, Lexiang adopts a "master Agent scheduling + specialized Agent handover" collaboration model: the master Agent is responsible for understanding user intent, breaking down complex tasks into executable sub-tasks, and deciding on serial or parallel execution based on task characteristics—serial for ensuring results, parallel for improving efficiency. Specialized agents each focus on their respective systems, completing the full business loop through a handover process.
Comments