By 2026, AI will no longer be a scarce resource. Visual recognition, multimodal large models, and agents will be readily available tools with low barriers to entry, almost accessible to everyone.

However, for industrial sites, none of the real problems have been solved—an anomaly is detected during inspection, but then what? Who evaluates it? Who handles it? Who follows up to ensure resolution?

The more tools available, the harder it becomes to achieve a closed loop. Capabilities that are repeatedly discussed require systematic clarification. Industrial scenarios are not lacking in 'visible' models; what they lack is a framework to integrate these models into business operations and create a closed loop—how anomalies are detected, how results are evaluated, and how dispositions are tracked.

Based on this, we turn our perspective to a practitioner deeply involved in industrial agents—Qiao Yuchuan, Director of Embodied Intelligence Solutions at SenseTime. He can address not just how to use a model but how AI can truly achieve a closed loop in industrial operations.

Conference Overview

With this question in mind, we will seek answers at the offline conference on October 23.

Hosted by OFweek, the "WAIE 2026 (10th) AI Industry Conference & AI Agent Industry Application Conference" will take place on October 23, 2026, at The Westin Shenzhen Nanshan.

This conference will focus on AI agents as the core theme, emphasizing both foundational technological innovation and practical implementation across various industries. It will provide an in-depth interpretation of national AI industry policies, the underlying architecture of general-purpose agents, and the pain points of industrial commercialization. The event will dissect the challenges of technological iteration and cross-industry integration in popular application areas, while sharing benchmark cases across the industry. Hundreds of government and enterprise leaders, research experts, R&D vendors, and core decision-makers from manufacturing, energy, government services, healthcare, education, finance, and commerce will gather for in-depth discussions on technological innovation, industrial chain collaboration, and industry-academia-research transformation. The goal is to connect government, enterprise, industry, academia, and research resources, deliver actionable agent solutions, showcase innovative products and technical solutions, help enterprises enhance their brand influence, expand collaboration opportunities, and comprehensively accelerate the large-scale implementation of AI agents across various industries.

Guests and Keynote Speeches

Recently, Qiao Yuchuan, Director of Embodied Intelligence Solutions at SenseTime, confirmed his attendance at the conference and will deliver a keynote speech titled "Collaborative Industrial Business Closed-Loop Practices of Multimodality and Agents."

In this speech, he will move beyond conceptual discussions of AI implementation in industry and address real pain points in industrial inspections, quality testing, and safety management. He will dissect how visual small models, multimodal large models, and agents are synergized to create a business closed loop covering anomaly detection, intelligent evaluation, task disposition, and result feedback. He will also explain on-site how recognition results are transformed into evidence-based analytical reports, actionable disposition orders, and traceable improvement records.

Keynote Speech: Collaborative Industrial Business Closed-Loop Practices of Multimodality and Agents

SenseTime has long invested in original research on multimodal, multi-task general artificial intelligence capabilities, building the "SenseCore" new-generation AI infrastructure and the "SenseNova" large model system. Its business covers four major sectors: smart commerce, smart cities, smart living, and intelligent vehicles. For industrial inspections, quality testing, and safety management, SenseTime integrates visual small models, multimodal large models, and agents into a single business chain, enabling AI to evolve from "single-point recognition" to "business closed loops."

Speech Highlights

01 From 'Visible' to 'Manageable'

Real pain points in industrial inspections, quality testing, and safety management: Detecting anomalies is just the first step; evaluation, disposition, and tracking are the truly challenging parts.

02 How Small Models, Large Models, and Agents Collaborate

Visual small models handle 'fast and accurate' anomaly detection, multimodal large models handle 'comprehensible' verification and evaluation, and agents integrate conclusions into business processes, each fulfilling its role.

03 Transforming Recognition Results into Actionable Closed Loops

How on-site information is linked to enterprise knowledge, business rules, and management systems to generate evidence-based analytical reports, actionable disposition orders, and traceable improvement records, with disposition feedback used for continuous optimization.

04 Live Demonstration: From Generating Reports to Initiating Work Orders

Through live demonstrations combining visual recognition, large model verification, and report generation and work order initiation by XiaoHuanXiong, the speech will explain how multimodal collaboration supports business processes and share implementation ideas and value validation methods for industrial agents.

Guest Introduction

Qiao Yuchuan

Director of Embodied Intelligence Solutions, SenseTime

◆ Currently serves as Director of Embodied Intelligence Solutions at SenseTime, focusing on the practical implementation of multimodal large models and agents in industrial scenarios

◆ Formerly a senior product manager at Huawei, Digital China, and DingTalk, with a career spanning large tech companies, internet platforms, and cutting-edge tech startups

◆ As a co-founder of a startup in the AI hardware and embodied intelligence space, successfully led two AI products to top industry sales rankings

◆ Possesses extensive experience in AI product commercialization and solution implementation

Conference Information