Li Yanhong Proposes the "Intelligent Organization," but Baidu Still Lacks an Entry Point
At Baidu World 2026, Li Yanhong called the "intelligent organization" the next act of AI, and proposed three principles: breaking down job boundaries, accumulating tacit knowledge, and designing closed-loop processes. Baidu's internal daily active agents have grown from 5,000 to 37,000, and weekly effective tasks have increased from 84,000 to over 1 million, but products such as Baidu Cloud, Dazi, Kuku, and Yijing remain fragmented and have yet to form a market-wide entry point. In September 2026, Baidu Dazi had about 9.77 million monthly active users, less than one-third of Tencent WorkBuddy's 32.353 million.
Body · AI translation
**Key points**
**1. The "intelligent organization" judgment is accurate, but Baidu did not originate it**
Microsoft already proposed the "frontier firm" in 2025, and major domestic tech companies are also changing their organizations; Baidu is mostly summarizing the proposition, not redefining productivity.
**2. Baidu has full-stack capabilities, but lacks a market-wide entry point**
Chip-cloud-model-agent, internal agents and the Token Exchange already have a foundation, but products such as the cloud, Dazi, Kuku and Yijing are scattered and have yet to form a closed loop that continuously draws in users and workflows.
**3. It lags in both office and content entry-point competition**
Tencent, Alibaba and ByteDance have already reorganized around their main entry points; Dazi has about 9.77 million MAU, ranking second and less than one-third of WorkBuddy's 32.353 million; Kling's revenue in Q2 2026 exceeded 850 million yuan, while Yijing has yet to prove itself at a comparable scale.
**4. Financial pressure forces Baidu to concentrate resources on star products**
As of the end of June 2026, cash and investments stood at 283.1 billion yuan, but Q2 free cash flow was about -7.95 billion yuan, R&D fell 10% and online marketing fell 19%; old businesses are under pressure and new businesses consume capital, so it can no longer spread its efforts evenly. 2026-11-14 08:41 Does the "intelligent organization" still need Baidu? Chaoyong AI Quick Look
English text is being prepared.
This judgment is not wrong; it may even be the most accurate summary of this year's organizational changes at major tech companies.
But what Baidu needs to answer is not "will the intelligent organization arrive," but "why should Baidu define it."
Seeing the future and owning the future are two different things.
Baidu has not lacked the former ability in the past; what it has never fully proven is the latter.
Robin Li summarized a future that is already happening
Robin Li proposed three evolutionary principles for the "intelligent organization": break down job boundaries, precipitate tacit knowledge, and design closed-loop processes.
Jobs are the unit of the industrial age; tasks are the unit of the intelligent age. As more and more execution work can be handed to agents, the organization will no longer be just a combination of people and reporting lines, but a network operating around tasks.
He further judged that jobs will not disappear in the future, but will converge into two types: Builders, responsible for creating new things, and Sellers, responsible for communication, connection and building trust.
Baidu also presented internal practice.
Robin Li disclosed that over the past year, Baidu's internal daily active agents grew from 5,000 to 37,000, and the effective tasks completed by agents each week grew from 84,000 to more than 1 million; Baidu has built 4,300 Workspaces and accumulated more than 20,000 organizational Skills.
So the "intelligent organization" is not an empty slogan.
Baidu is indeed trying to turn employee experience into reusable organizational assets, let agents enter workflows, and then continuously optimize through feedback.
The problem is that none of this is a new continent first discovered by Baidu.
In its annual Work Trend Index report released in April 2025, Microsoft already proposed the "frontier firm," discussing mixed teams of human employees and agents, agents entering business processes, and how enterprises can precipitate AI usage experience into organizational capability.
Microsoft's 2025 Work Trend Index
Robin Li's contribution is to summarize these ongoing changes into a complete proposition, not to redefine productivity.
Changes in productivity happen in the daily work where tasks are rewritten, knowledge is precipitated and processes begin to close the loop, not at the moment a concept is named.
Calling what is happening in the industry the "next act" does not mean one has already gained control of this act.
Baidu has a full-stack system, but lacks a main axis running through the market
In terms of system completeness, Baidu remains one of the few players among major domestic tech companies.
From Kunlun chips, Baidu AI Cloud and ERNIE models to Baidu Dazi, Kuku, Baidu Yijing and Famou, Baidu has built a four-layer "chip-cloud-model-agent" architecture. At this conference, the ERNIE 5.5 Flash preview, the Tianchi 1024-card super node and upgrades to four agents pushed this system one step further.
The newly launched "Token Exchange" attempts to build a large-model trading platform, using authenticity verification and credit reporting to address issues of model provenance, service quality and invocation stability. Baidu disclosed that one month after launch, it had more than 100 supplier resources onboarded, with average daily Token invocations exceeding 700 billion.
These products at least show that Baidu is not using an architecture diagram to talk about intelligent organizations. It has a technology foundation, internal scenarios, and a batch of applications that are looking for a market.
But full-stack answers "what can I provide"; a market entry point answers "why must customers work through me every day."
These are two different questions.
It is not that Baidu has no strong businesses.
Baidu's Q2 2026 financial report shows that AI cloud infrastructure revenue grew 50% year on year, and GPU cloud revenue grew 283% year on year; Apollo Go, digital humans and other segmented businesses also have their own accumulations...
One cannot simply say that Baidu "has nothing it can put forward to fight."
Its real problem is that these businesses have not yet formed a main axis that pulls the "intelligent organization."
The cloud provides computing power, the Token Hub connects models, Dazi targets office work, Kuku handles documents and content, Yijing creates digital humans, and Famo is aimed at industry decision-making. Each piece has value, but there is not yet a strong enough product that continuously draws users, data, tasks and workflows into the same closed loop.
What Baidu lacks most is precisely this axis.
The outside world often describes Baidu as "rising early but arriving at the late market," and applied to the matter of the "intelligent organization," this points not to Baidu failing to understand the trend, but to the gap between its technology system and market results: it laid out its plans very early and has many products, but in the end it has not formed a sufficiently strong entry point.
An architecture can be drawn in one go at a conference, but an entry point can only be won inch by inch through a star product. Without an entry point, full-stack capability is just a set of parts waiting for customers to assemble themselves; with an entry point, it can become an organizational system that is hard for others to replace.
Rivals have already reorganized around entry points, while Baidu is still looking for a flagship product
From July to August 2026, Tencent, Alibaba and ByteDance successively adjusted their AI office businesses.
Tencent merged QClaw into the department where WorkBuddy sits; Alibaba merged and reorganized three product lines—QoderWork, Wukong and MuleRun—into the Qwen Office business unit; ByteDance merged the Feishu product team into the Doubao system, then brought TRAE and Coze into Doubao, and subsequently released "Doubao Work."
These adjustments may not all be correct, but they send a common signal: competition in intelligent organizations is no longer just about how many Agents are released, but about who dares to end internal horse racing and concentrate resources on a main entry point validated by the market.
Baidu's Dazi is also doing similar integration and has achieved real growth, and it should not be written off as "not ranking near the front."
According to QuestMobile data, in September 2026, Baidu Dazi had about 9.77 million monthly active users, ranking second among domestic AI office intelligent agents; but the first place, Tencent WorkBuddy, had 32.353 million, and the third place, Qwen Office, had 7.905 million.
This ranking precisely reveals Baidu's current position: it is already qualified to sit at the table, but does not yet control the game.
There is still a distance between second place and an entry point. Dazi's monthly active users are less than one-third of WorkBuddy's, while Tencent can connect office intelligent agents to WeCom and WeChat's relationship chains, and Alibaba and ByteDance are also concentrating their respective products, data and organizational processes toward main entry points.
The situation in video and content production illustrates the problem even more clearly.
Baidu Yijing has its own position in the digital human niche market and cannot simply be conflated with general video generation products. But if the view is widened to the content production entry point, the market has continuously voted for Jimeng and Kling with users and revenue. Kuaishou disclosed that Kling AI's revenue in the second quarter of 2026 exceeded 850 million yuan, up more than 200% year on year.
At least judging from public data, Yijing has not yet produced proof of users or revenue on the same scale.
This is Baidu's most realistic shortcoming: it is not that it has made no effort, nor that no products have made the rankings, but that it lacks a flagship product capable of pulling full-stack capability and establishing market mindshare.
An intelligent organization needs a high-frequency, trustworthy entry point that can continuously accumulate organizational knowledge. What Baidu is offering now is still a set of products, not this definitive answer.
Of course, the intelligent organization needs Baidu.
It needs Baidu's chips, cloud, models, intelligent agents and internal practices, and it also needs an established technology company to participate in exploration. But participating in a productivity transformation and becoming the standard-bearer of this transformation are two completely different market positions.
For Baidu to carry the flag, it cannot rely on "I was the first to propose this concept," but should rely on "the product that the most enterprises are willing to entrust their real work to."
Robin Li is not unaware of where the problem lies. In 2025, in an internal Baidu speech, he directly asked "why can't Baidu win battles," and attributed the reason to "lack of focus."
Looking again at the "intelligent organization" today, this statement still applies.
It is not that Baidu has no money.
According to Baidu's Q2 2026 financial report, as of the end of June 2026, its total cash and investments were 283.1 billion yuan. But in the same quarter, Baidu's capital expenditure reached 11.39 billion yuan, operating cash flow was 3.436 billion yuan, and free cash flow was about negative 7.95 billion yuan; R&D expenses were 4.607 billion yuan, down about 10% year on year. Online marketing revenue fell 19% year on year, and the traditional cash cow continued to come under pressure.
Baidu Q2 2026 financial report
So the problem is not simply "insufficient financial strength," but that old businesses are contracting and new businesses are more capital-hungry, and Baidu can no longer afford to continue spreading its efforts evenly.
The more this is the case, the more it needs one product to concentrate resources and break through the market, rather than adding another layer of grand conceptual packaging.
Pony Ma said in Tencent's internal livestream on November 11 for its 28th anniversary that the AI transformation is a "life-and-death transformation to adapt to new battlefields and new ways of fighting." He stressed that this is no longer simply about cost reduction, but that both people and organizations must adjust, core businesses must improve efficiency, and at the same time forces must be split to explore new AI Native battlefields.
This sense of urgency applies equally to Baidu, and even more so.
Tencent has already been concentrating resources around WorkBuddy, and ByteDance and Alibaba have also each found their own main product axis. If Baidu still puts its main energy into defining the next act rather than winning the most specific users and scenarios in this act, it will be difficult for it to change how the outside world judges it.
The way to tear off the label is not to be the first to propose another new concept, but to make one product that others cannot avoid.
If the star product and market entry point are still absent, then the more correct the judgment of "intelligent organization" is, the more likely it will once again prove: Baidu saw the era, but the era handed the entry point to someone else. AI industry signal channel: Frontier Technology. This content is published with the author's authorization, and the views represent only the author himself and do not represent Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is in the Huxiu APP[ ]]]