Boao现场｜Baidu Smart Cloud Releases Top 10 Intelligent Agent Cases: Intelligent Agents Reshape the Future of Industry

On March 26, at the 2026 Boao Forum for Asia, Baidu Intelligent Cloud officially released the “Top 10 Enterprise-Level AI Agent Cases of 2025,” showcasing Baidu Intelligent Cloud’s end-to-end, fully self-developed capabilities based on its “Chip-and-Cloud Molds” platform in recent years, helping enterprises expand AI agent application scenarios.

Unlike the technical showmanship in labs, all the agent cases unveiled in this release come from the front line of industry. They are not only reliable, professional Agents, but also “digital employees” that help businesses reduce costs and improve efficiency. At the event, Cao Haitao, General Manager of Baidu Intelligent Cloud’s market department, said that these cases represent the wave of AI agents accelerating into real-world deployment across industries. The direction of industrial development is becoming clearer, and large models have entered a new stage where value is defined by applications. In the future, Baidu will work with customers, partners, and developers to step into the most complex real scenarios and turn intelligence into tangible value.

The first case shown that day came from the power grid inspection AI agent of the Beijing Humanoid Robot Innovation Center—“Tiangong.” With the help of Baidu Intelligent Cloud, Tiangong became the first humanoid robot to independently complete operations and inspections in the power industry. For example, at substations in remote areas in the southwest, in the past, staff needed to travel over mountains and ridges to carry out inspections and test equipment, which was very labor-intensive. Now, the Tiangong robot enables outdoor inspections and checks the exterior of substations; at the same time, it can also perform operations of power equipment inside substations. This also lays a good foundation for the next step of enabling embodied inspections in high-altitude, unattended areas.

In the field of automotive industrial design, Altcar Technology Co., Ltd., which has more than 20 years of experience, previously faced a bottleneck: wind-resistance testing took 10 hours. Based on Baidu’s “Faimou” intelligent agent, Altcar wrote physical constraints from aerodynamics and fluid mechanics, as well as design features such as the vehicle body styling, into the algorithm, enabling wind-resistance evaluation at a minute-level. Today, the same type of vehicle model can directly use the same Faimou algorithm model to conduct wind-resistance evaluation—and it can be completed at the minute level. Altcar’s designers only need 2 days to achieve results that previously required multiple engineers working together for 3 months, greatly improving production efficiency.

AI is continuously helping tackle hard problems in real-world scenarios. Previously, Beijing University of Technology faced challenges in the design of a micro gas chromatograph for a space station. Using traditional methods, they spent more than a year, but the results were still not ideal. After introducing Baidu’s self-evolving intelligent agent “Faimou,” the “configuration” problem of the chromatographic column was abstracted into a mathematical modeling and optimization problem, allowing the agent to automatically compute and evolve. The outcome was that the testing error was reduced by 8% compared with the manual solution, and the configuration designed by Faimou can help air flow more evenly; the separation efficiency of air molecules has been greatly improved.

In the consumer electronics service side, OPPO’s after-sales customer service team deals with a huge volume of colloquial inquiries every day. Together with Baidu Intelligent Cloud, it rolled out a customer service intelligent agent that directly connects to OPPO’s accumulated service experience of more than 100,000 service articles, enabling AI to understand “charging port” instead of the rigid term “tail plug.” The efficiency for handling frequent issues improved by 1x, and in small-language scenarios such as Vietnamese and Indonesian, the results are also excellent—turning service capability from reliance on individual experience into a stable, replicable corporate asset.

Based on Baidu Intelligent Cloud’s AI infrastructure capabilities, the changes brought by intelligent agents are also taking place in key industries such as finance, transportation, and energy. China Eastern Airlines’ digital employee “Dongdong” has internalized core processes such as ticket booking, seat selection, and check-in, providing one-stop proactive services on the App. The State Grid assigned the generation of complex power supply plans to intelligent agents. Galaxy Securities’ trading intelligent agent can understand the industry’s “black talk,” assisting traders to improve the conversion efficiency from inquiry to order by 3x, with business scale doubling. Baixin Bank uses intelligent agents for risk control, letting the large model evolve risk features by itself, significantly improving risk identification capability. At Qingdao Port, a safety production intelligent agent enables hazards to be reported with “no omissions and no false reports,” and its AI capability covers the entire port; while in Ordos, a credit-control intelligent agent makes traffic lights “learn to think,” reducing average delays per vehicle in the main urban area by 18%. These intelligent agents across different industries collectively confirm a trend: intelligent agents are accelerating into high-value industrial application scenarios, injecting new vitality into the intelligent economy.

Behind these ten cases is strong support from Baidu’s end-to-end AI full-stack infrastructure. From chips, cloud, and models to intelligent agents, Baidu Intelligent Cloud achieves end-to-end coordinated optimization. Data shows that in 2025, Baidu Intelligent Cloud’s growth exceeded 30%, serving more than 80% of central state-owned enterprises, while holding the #1 market share in the fields of finance, automobiles, and embodied intelligence. Meanwhile, to make AI capabilities even more accessible, Baidu Intelligent Cloud has also recently built a complete tool ecosystem around the OpenClaw hotspot, launching on the consumer side including the cloud-based shrimp DuClaw, the world’s first mobile lobster application RedClaw, and “Xiaodu (little) shrimp” adapted for home scenarios. For enterprises, the domestic enterprise-level lobster product DuMate just released last week supports local deployment, and can easily handle scenarios such as automatic form filling and reimbursement approval, ensuring data security and controllability. In addition, Baidu also continuously enriches the Skills ecosystem through its own products. Baidu Search has already become the official Skill with the highest global download volume.