Add preferred source
Baidu AI Cloud announced in Shenzhen on September 20 that its general-purpose agent product Baidu Dazi saw user scale grow 9-fold month-over-month, deliverable demand rise 4.3-fold, and expert suite usage jump 7.3-fold. The company unveiled an enterprise mobile app, an agent development platform, and launched an ecosystem co-creation program. The enterprise edition adds project spaces, project memory, and handoff capabilities, and has connected specialized agents including Miaoda, Famou, Hogee, and Shengsuan, offering over 13 industry suites. The development platform opens Agent API and Skill configuration, supporting public cloud, dedicated VPC, and on-premises private deployment. On the hardware front, Baidu introduced a standardized API, with devices from Xiaodu and Yinshi Technology already integrated. Baidu AI Cloud previously introduced the “Industrial Agent Operating System” concept, leveraging full-stack “chip-cloud-model-agent” capabilities to orchestrate general, specialized, and custom agents within enterprise workflows, aiming to build a growth flywheel from compute to applications.
Key Elements
Baidu AI Cloud announced at its Super Agent Conference in Shenzhen on September 20 that its general-purpose agent product Baidu Dazi saw user scale grow 9-fold month-over-month, deliverable demand rise 4.3-fold, and expert suite usage jump 7.3-fold. The company simultaneously launched an enterprise mobile app, an agent development platform, and an ecosystem co-creation program, marking a shift in Baidu’s industrial agent strategy from standalone tools to systematic deployment.
Shen Dou, Executive Vice President of Baidu and President of Baidu AI Cloud, said at the event that enterprise intelligence requires general-purpose agents, specialized agents, and custom agents to work in concert. Baidu Dazi, as a general-purpose agent, handles daily operational tasks while connecting specialized agents, enterprise custom agents, and existing business systems, pushing agents beyond conversational interaction toward task execution and business delivery.
Just four days earlier, on September 16, Baidu AI Cloud first introduced the “Industrial Agent Operating System” concept at the 2026 Intelligent Economy Forum. Shen Dou categorized enterprise tasks into three groups: 20% daily operations, 30% specialized domain tasks, and 50% core tasks that determine business outcomes. He argued that these task capabilities will gradually be internalized by corresponding agents over time.
Enterprise Edition Strengthens Multi-Person Collaboration and Mobile Work
For team collaboration scenarios, Baidu Dazi Enterprise Edition adds three capabilities: project spaces, project memory, and share-and-handoff. Teams can centralize members, tasks, files, and related information around shared goals in project spaces; project memory continuously accumulates business rules, key decisions, and project outcomes for later retrieval; share-and-handoff enables work content and results to flow between members, reducing redundant communication and copy-paste work.
The newly released enterprise mobile app supports phone-to-PC handoff. Users can initiate tasks on mobile, monitor execution progress, receive result notifications, and access enterprise systems to continue work. This extends Baidu Dazi from desktop to mobile scenarios, aiming to cover enterprise users’ around-the-clock work rhythms.
For specialized tasks, Baidu Dazi has connected professional agents including Miaoda, Famou, Hogee, and Shengsuan, and launched over 13 industry suites covering legal, financial, automotive, government, education, and marketing sectors. Each suite integrates industry knowledge, specialized data, Skills, and workflows, allowing enterprises to invoke relevant capabilities for specific tasks without training models from scratch. At the event, Baidu Dazi also jointly launched a marketing insights suite with Hogee, supporting sentiment monitoring, competitor tracking, user review analysis, and market opportunity research.
Developer Platform Opens Underlying Capabilities
For enterprise customization needs, Baidu Dazi launched an agent development platform, opening Agent API, tool and Skill configuration, and agent runtime capabilities. Enterprises can integrate internal knowledge, data, permissions, systems, and business processes to develop agents tailored to their needs. The platform uses the same Harness execution engine and memory architecture as Baidu Dazi, supporting runtime monitoring, security management, and deployment options including public cloud, dedicated VPC, and on-premises private deployment.
Baidu Dazi also kicked off an ecosystem co-creation program, opening connectors, specialized Skills, and scenario co-creation capabilities to partners. Software, data, and professional services can connect to Baidu Dazi through connectors; industry expertise and tools can be distilled into reusable Skills, which can then be combined into agent solutions for specific business needs.
Beyond the software and services ecosystem, Baidu Dazi is expanding its hardware ecosystem with a standardized hardware API that enables capability parity, multi-device synchronization, and memory synchronization, along with a complete integration toolchain including standard protocols, debugging tools, and compatibility certification. Partners can gain full Dazi capabilities through a lightweight three-step integration process. Xiaodu home devices and embodied intelligence devices from Yinshi Technology are already connected to the standard API.
The Underlying Logic of Industrial Agents
Behind this dense slate of releases is a full-stack strategy spanning from compute to applications. Baidu’s Q2 2026 earnings showed Baidu AI Cloud infrastructure revenue reached 7.3 billion yuan (approximately $1.1 billion), up 50% year-over-year; GPU cloud revenue grew 283% year-over-year, accelerating from 184% growth in the previous quarter. Baidu noted that GPU cloud revenue growth reflects sustained market demand for public cloud AI compute.
Baidu AI Cloud frames this system as full-stack “chip-cloud-model-agent” capabilities: underlying chips provide compute power, AI Cloud handles compute and resource scheduling, models deliver intelligence, and ultimately three types of agents bring these capabilities into specific business scenarios. At the compute layer, Kunlun Core has accumulated over a decade of development, with the third-generation P800 now deployed at scale across internet finance, manufacturing, automotive, and telecom operator sectors. At the Agent Infra layer, Baidu AI Cloud offers over 150 state-of-the-art models, long-chain reasoning optimization, built-in sandboxing, and risk control capabilities, focusing on elasticity, security, and transparency in agent operations.
Baidu AI Cloud disclosed that it serves over 800 financial institutions, covering all systemically important banks in China, and has ranked first in China’s financial large model market share for two consecutive years; in automotive, it covers mainstream Chinese automakers and maintains leadership in autonomous driving R&D solutions; in embodied intelligence, it serves over 50 related enterprises.
These industrial customers bring not standardized office needs but massive volumes of real business processes. For agents to enter these processes, they must solve connectivity challenges across data, knowledge, permissions, tools, models, and compute. Baidu AI Cloud believes that while enterprises previously purchased standalone software products, more capabilities will increasingly be packaged as Skills and invoked on demand by agents. As agents generate more real business data, that data can be distilled into enterprise knowledge and capabilities, driving agent iteration; as agent deployment scales, it generates more model calls and compute demand, creating a growth flywheel from underlying infrastructure to enterprise operations.
From a competitive landscape perspective, AI office productivity has become the most fiercely contested To B battleground among Chinese tech giants, with multiple vendors consolidating product lines to launch office agents. But Baidu AI Cloud’s assessment is that office work represents only a portion of enterprise operations—most of the critical tasks that determine business outcomes occur outside office software. R&D, production, risk control, and design involve large volumes of complex, highly specialized tasks. This is why Baidu chose the “Industrial Agent Operating System” as its differentiation point.
Once added, BigGo Finance appears first in Google Search Top Stories, so you get the broadest, most up-to-the-minute, and most comprehensive global financial news first.
