AICon 2026 Shenzhen Concludes: AI Implementation Enters Deep Water, Focusing on Engineering Practice and Industrial Value

From August 21 to 22, the 2026 AICon Global AI Development and Application Conference · Shenzhen, hosted by InfoQ China under Geekbang Technology, was held, attracting nearly 1,000 developers, technical experts, and industry practitioners. The conference focused on AI engineering implementation, Agent innovation practices, large model efficiency engineering, and full-process software development restructuring, inviting over 70 experts from 40+ enterprises and research institutions to share insights.

In the opening address, Geekbang Technology's Editor-in-Chief Zhao Yuying proposed that AI transformation is shifting from tool deployment to restructuring production relations. The real gap between enterprises is not whether they use AI, but whether AI can enter the business loop and reshape organization and talent. She pointed out that current AI application implementation shows value differentiation after the 'hundred-regiment war,' with a common pain point of emphasizing quantity over efficiency.

Zhao Yuying emphasized that the key to talent transformation lies in team-based capability building. Large and medium-sized enterprises are actively promoting younger AI leaders, and the value of FDE delivery talent and full-stack engineers is rising. However, the FDE role should not be mythologized; capability building should be team-oriented rather than individual-oriented.

In his speech, Qunqing Intelligence co-founder and CEO Wu Zheming stated that the key to judging whether industrial embodied intelligence is viable is not whether a robot can complete a demonstration, but whether it can independently undertake labor with clear economic value. If a robot is more expensive and slower than a human, and still requires constant supervision, it has not truly become a productivity tool.

Qunqing Intelligence equips industrial robotic arms with 3D vision and other sensing devices, building an industrial intelligent agent system composed of autonomous perception, autonomous planning, edge execution, and closed-loop learning. They also independently developed the Lapis Lazuli Physical AI architecture, integrating physical constraints, process mechanisms, expert knowledge, and real production data.

Currently, their products have been applied on a large scale in industries such as transportation, energy equipment, and engineering machinery, covering over 30 enterprises with a customer repurchase rate of 90%. They have accumulated over 300TB of real welding process data, with a typical project investment payback period of about 0.5 to 1.5 years, and a single factory has completed over 100,000 welds.