Newshot

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

AI summary

Ten years ago it was about AI; ten years from now it will be about intelligent organizations. Employees are already using agents to get work done and even building their own small systems, while company KPIs, approval chains, and reporting cadences still run the way they did three years ago. A recent Deloitte survey found that nearly three-quarters of U.S. executives expect about half of business processes to be redesigned around AI agents in the next four years, yet only one in five leaders says their company is ready for that change.

Translation(model translation, not the source's own text)

Title: AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

In recent conversations with people in the AI circle, one point comes up universally: people have already accepted AI for efficiency gains, but enterprises are not yet ready. (Leiphone (WeChat public account: Leiphone) Leiphone Leiphone)

On one side, employees are already using agents to get work done, and some have even started building their own small systems; on the other side, the company's KPIs, approval chains, and reporting cadence are still running the way they did three years ago.

A recent Deloitte survey shows that nearly three-quarters of U.S. enterprise executives expect that over the next four years, about half of business processes will be redesigned around AI Agents; 61% believe that in the future most Agents will be highly autonomous, with humans only needing to supervise from the sidelines.

It seems that AI taking over all business is the general trend, but when asked whether their enterprises are ready for this change, only one in five leaders gave an affirmative answer.

At today's Baidu World 2026, Baidu founder 李彦宏 also touched on this contradiction. His judgment: AI is not just a technological revolution, but an organizational revolution; the next act of artificial intelligence is the intelligent organization.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

What enterprises need to change next is not just the tools in employees' hands, but also the way the organization itself operates.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

AI Has Entered the Company, But Not Completely

To restructure how an organization operates, the first step is to break down a common assumption:

A high concentration of AI in a company does not mean AI transformation has already happened.

Employees use Agents to write code, do research, and generate proposals; business units get a few pilots running; management then writes the number of agents, Token consumption, and usage rates into reports—it looks like a company's intelligentization is advancing across the board.

But individual efficiency does not automatically raise organizational productivity.

For example, requirements analysis that used to take a product manager two days may now take half a day; R&D uses Agents to quickly write first drafts of code, and sales can generate a research report before visiting a client.

But requirements still have to go through review, approval, and scheduling; cross-department collaboration still relies on meetings and layer-by-layer synchronization; and final results still have to wait for the original reporting cycle to be seen. After going around in a big circle, employees' personal time is reduced, but the enterprise's overall organizational capability has not improved noticeably.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

What Microsoft's "2026 Work Trend Index Annual Report" presents is precisely this enterprise AI transformation paradox: employees have already begun using AI to improve personal efficiency, and teams are also trying new tools, but business processes, collaboration methods, and evaluation mechanisms still follow past models.

The advice from Boston Consulting Group (BCG) is even more direct: redesign the company for AI, rather than making AI adapt to the original company.

A European energy company mentioned in its research had once deployed a large number of AI pilots but did not obtain overall returns matching them. Not until the company shifted its focus from individual tools and local scenarios to reconstructing end-to-end customer processes did it achieve substantive business improvement.

Such cases show that between the success of AI pilots and the success of enterprise transformation, there is still a layer of organizational design in between.

If an enterprise merely adds tools to existing positions and inserts a few Agents into fixed processes, then AI can optimize local parts but can hardly change the whole.

It may even make existing bottlenecks more obvious: content is produced faster and faster, but decision-making does not speed up in sync; more and more code is generated, but requirements confirmation and delivery mechanisms remain the same.

When execution capability is rapidly amplified, how the organization allocates tasks, shares knowledge, and makes decisions instead becomes the new constraint.

Over the past few years, the main competition in the AI industry has centered on the technology supply side.

Model parameters, training compute, inference costs, and leaderboard rankings determine what kind of intelligence enterprises can obtain.

The industry once popularized the AI "five-layer cake," extending from energy, chips, and infrastructure all the way to models and applications, emphasizing precisely the complete system required for intelligence production. Without energy and compute, the upper-layer models and applications cannot be discussed at all.

Changes in model supply are also beginning to show on the transaction side. One month after Ciyuan launched, the number of onboarded supply resources has exceeded 100, average daily calls have surpassed 50 million, and average daily Token calls have exceeded 700 billion.

More noteworthy than the scale itself is that model trading is moving from private connections between enterprises and vendors toward an independent industrial segment; as this segment takes shape, the rule provider also gradually takes on the attributes of a public good.

But on the enterprise side: when this intelligence can already be purchased, called, and deployed, what determines how much value it can ultimately generate?

李彦宏 turned his gaze downstream. At the 2026 Baidu World Conference, 李彦宏 proposed that "the next act of AI development is the intelligent organization."

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

He proposed that the upstream of AI technology is energy, and the downstream is organizational capability. The former determines how intelligence is produced, and the latter determines how intelligence enters real business and is converted into business results.

These two judgments are not contradictory; rather, they answer questions at both ends of the AI industry chain. For enterprises, even if the "electricity" provided upstream is sufficient, if the interior is still an old production system, the added intelligence may not necessarily become added profit.

李彦宏 compared AI to a new electric motor and the intelligent organization to a new factory building.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

He judges that an important reason AI has not produced the expected economic benefits is that enterprises have only replaced the motor but have not yet redesigned the factory.

Models, compute, and Agents can provide stronger execution capability, and only when task allocation, knowledge flow, and decision-making processes change accordingly can this capability possibly move from employees' personal desktops into the enterprise's complete business chain.

The focus of AI competition has thus begun to shift further from technology supply toward value realization.

What enterprises need to answer is no longer just "what models I used and how many Agents I deployed," but whether these agents have actually changed the company?

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

Workers Improve Efficiency with AI, But What About the Company?

Microsoft's CEO used to say often: AI is not a technology, but the future form of the enterprise.

Once you understand this sentence, you can actually understand the "intelligent organization" proposed by 李彦宏.

Frankly speaking, if you look only at the literal meaning, it is easy to understand it as a new type of company composed of people and Agents, or to treat it as another way of saying "AI-native enterprise."

But what 李彦宏 wants to discuss is not equipping every position with an AI assistant, but redesigning the way the enterprise works: who completes tasks, how experience is accumulated, and how processes continuously adjust based on results.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

Traditional enterprises operate through professional division of labor, and a complete task often requires relay completion by multiple positions such as product, design, R&D, and operations.

In the AI era, agents begin to take on the specific execution work within it, while people are responsible for setting goals, scheduling resources, and making key decisions.

The organization no longer allocates work entirely around fixed positions, but can flexibly combine people and agents according to task needs.

The practice of Baidu's Dazi team provides an example. A member who originally had a background in quantum computing research joined Baidu and entered product and enterprise scenarios almost from scratch; with the help of Baidu Dazi, he simultaneously took on work in product, R&D, and project management. From analyzing problems and forming proposals to advancing execution and verifying results, he was no longer limited by a single fixed position.

New product staff can rely on Dazi to quickly run through the entire process from requirements to prototype; the team iterated more than 160 times in half a year, with 90% of code and more than 50% of communication materials completed by Dazi.

In this case, what is truly broken is not only the rule of "who should do what," but also the limits on the flow of knowledge and experience.

The Baidu Digital Human team's push for roles such as product, R&D, design, and testing to transform into Builders is also an extension of the same logic.

The point is not to make everyone unconditionally take on more work, but to reduce repeated handoffs of tasks between different roles, so that people who can define goals also have the opportunity to use AI to reach the final result.

But if only the way tasks are executed changes, the intelligent organization can only be considered half completed; the other half still depends on how the organization makes decisions.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

The knowledge that is truly valuable to an enterprise is employees' practical experience: for example, the judgment methods veteran employees use to handle faults, salespeople's experience in identifying customer needs, and designers' grasp of aesthetics and standards.

This knowledge is difficult to standardize and write into documents, and is also easily lost as personnel move.

But AI provides a new way of accumulating it: converting expert experience, business rules, and operating methods into knowledge bases, Skills, and capabilities that agents can call.

Baidu's cooperation with China Southern Power Grid involves this process. For distribution network fault analysis, traditional manual analysis takes more than 60 minutes per instance and is highly dependent on expert experience.

After Baidu Dazi was implemented in the Southern Power Grid general dispatch center, by learning from tens of thousands of historical fault handling data records, it accumulated senior experts' experience into reusable professional knowledge, and then linked fault records, switch signals, and line topology to form a complete evidence chain. At present, its fault analysis accuracy rate has reached over 70%.

AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?
AI Has Entered the Enterprise, So Why Hasn't the Organization Evolved?

Baidu MEG's design team demonstrated another way of accumulating knowledge.

Values written like {{NAME_ab12cd34}} are placeholders for real values. Use them exactly as written, in replies and in tool calls alike: they are filled in with the real values before anything runs. Don't ask to verify them, and don't replace them with examples.

The team turned senior designers' aesthetic experience and design specifications into reusable AI Skills, so that AI not only generates images according to instructions, but can also call on the design methods the team has accumulated and take part in content production. Today, AI tools already handle more than 50% of content supply, and team efficiency has improved by 100%–200%.

From power grid faults to design content production, the logic behind the two scenarios is quite similar: turning work methods that originally depended on individual experience into capabilities the organization can call on repeatedly.

Of course, as the business environment changes, experience can also become outdated. That is why task results need to flow back into the process.

This is the design process loop mentioned at the end.

Traditional processes are usually designed in advance by people; when problems arise during execution, people then review and correct them.

After each task is completed, the system can measure the result against preset goals, identify problems in execution, and then use the feedback to adjust the next round of tasks. In this way, the process is no longer a fixed set of operating rules, but a system that can be continuously corrected as the business runs.

李彦宏 summarized this as "measurable + Lopping = continuous evolution."

"Measurable" means task results need clear evaluation criteria, while Lopping means execution, feedback and optimization must form a continuous cycle.

When more and more tasks can complete such a loop, the organization no longer relies solely on managers to push things forward, but can continuously accumulate experience and correct processes in its daily operations.

To support such a loop, a single model API is obviously not enough.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

Agents need to understand the company's business goals, call on data, knowledge and tools from different systems, and be able to keep executing in long-chain tasks, accept evaluation and adjust based on feedback. This places new demands on the underlying infrastructure.

This is also the significance of the AI cloud and Agent Infra that Shen Dou emphasized at the 2026 Baidu World Conference.

Following Baidu's approach: the underlying Baige AI Infra provides computing power support, while the upper-level Qianfan Agent Infra provides long-context management, persistent memory, tool calling, sub-agent scheduling, evaluation feedback and a runtime environment, enabling large-scale, long-cycle tasks to truly complete the loop.

Placed in the full-stack structure of "chip—cloud—model—agent," this supporting relationship becomes clearer: chips provide the computing foundation, the cloud provides the infrastructure for agent operation and scheduling, models provide understanding, reasoning and generation capabilities, and agents enter specific tasks and business scenarios.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

From positions to tasks, from individual experience to organizational intelligence, and then to the continuous iteration of processes, what the intelligent organization changes is not the work efficiency of a certain position, but the underlying logic of enterprise production and collaboration.

What comes next is exactly how these principles become practice: truly taking part in business processes and solving real problems?

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

How does an intelligent organization grow inside a company?

To judge whether a company has entered the AI Native organization stage, you need to see whether AI is truly changing the entire internal operation.

Take Baidu's digital human innovation business department.

The team broke down the boundaries between product, R&D, design and testing roles, pushed all staff to transform into Builders, and based on the self-developed BiuBiuAgent, enabled some requirements to be delivered end-to-end by a single person.

In the third quarter alone, the team completed 564 product and R&D iterations, accumulating 130 Skills and 729 organizational assets.

As Loops continue to run, per capita throughput in product and R&D increased by 125.3%, and the delivery cycle was shortened by 28.8%. With the help of AI, a new campus hire completed in just one day a requirement that previously needed 5 roles and at least 10 days to deliver.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

Behind the rapidly growing numbers is AI truly "turning" within the enterprise organization.

In this way, employees can focus more energy on goal judgment and process organization, the Skills and organizational assets formed by teams during execution can be called on repeatedly, and an individual's experience no longer disappears when a task ends.

Another example is dodo. This AI assistant, built by a team of about a dozen people inside Baidu, covered 99% of regular employees seven months after launch and drove more than 300 enterprise platforms to complete AI-friendly transformation.

This also reveals a side of organizational intelligence that is easily overlooked: companies cannot just ask employees to learn to use AI; they also need existing systems, data and processes to be able to work with AI.

Otherwise, even if an agent can complete the task of a certain step, it will be hard to truly connect the business chain.

Over the past year, the number of daily active agents inside Baidu grew from about 5,000 to 37,000, and the number of effectively completed tasks per week rose from 84,000 to more than 1 million.

Beyond scale expansion, knowledge accumulation is happening at the same time: Baidu has built more than 4,300 workspaces in total and accumulated more than 20,000 company-level general skills, gradually turning experience that was originally scattered among individuals and teams into reusable organizational capabilities.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

If Baidu's practice shows changes in the way organizations collaborate, then the following cases test whether AI can enter real business.

At China Southern Power Grid, Baidu's partner solution was deployed at the Southern General Dispatch Center, forming "Tianxuan Partner" for distribution network fault analysis.

Guangxi Power Grid had 8,215 permanent faults in 2025; traditional manual analysis took more than 60 minutes per case and relied heavily on expert experience.

Tianxuan Partner learned from tens of thousands of historical fault-handling records, distilling senior experts' judgment experience into reusable knowledge, and then linked fault records, switch signals and line topology into an evidence chain to assist fault analysis.

According to customer feedback, fault analysis efficiency improved 12-fold, analysis accuracy reached more than 80%, and expert experience reuse was achieved; it has now been rolled out in 14 cities and 112 counties in Guangxi.

The value of this kind of application is not just making fault analysis faster, but turning experience originally held by a few experts into professional capability that the system can continuously call on.

Haigao Laser's practice brought AI to a more complex decision-making step on the manufacturing floor.

On the production line, multiple six-axis robotic arms need to work together to cut the same workpiece. In the past, process engineers mainly relied on experience to allocate trajectory blocks and determine execution order; but as the number of robotic arms, processing tasks and process constraints increase, local arrangements can easily affect overall efficiency.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

Baidu Famou continuously optimizes algorithms around motion time prediction, multi-machine collision avoidance and task allocation scheduling. While satisfying constraints such as robotic arm reachability, process dependencies and no collision, it compressed workpiece programming time from 8 hours to 10 minutes, solving a bottleneck problem in the industry.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

From power grid fault analysis to robotic arm collaborative scheduling, the two cases point to the same change: for AI to enter industry, it must understand industry knowledge, adapt to business constraints, and be tested by results in real tasks.

Of course, an intelligent organization does not mean that the people in the organization will become unimportant.

At the conference, a conversation between digital human Xijiajia and Robin happened to touch on this question.

When Xijiajia asked whether she could become CEO, Robin's answer was no.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

In his view, a CEO needs to understand the company's mission and vision; AI can become the CEO's thinking partner, helping improve efficiency and provide inspiration, but there is no need to rush to replace the CEO.

When the topic further turned to "if AI permeates every corner of the organization, what is still the core part of the organization," Robin's answer was: people.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

We often say AI cannot replace humans, but in practical applications there is still worry about being replaced; this conversation instead distinguished AI's execution power from organizational direction.

AI can continue to take on specific work, participate in analysis and provide suggestions, but the value of the company's existence, and where it goes in the future, are ultimately decided by "people."

This is somewhat like the "DRI (Directly Responsible Individual)" concept Apple proposed early on.

At Apple, who should be responsible for what has always been clear. Simply put, AI can help you make decisions, but the final result and value judgment must have someone responsible.

In other words, an intelligent organization does not remove people from the process, but redistributes the responsibilities borne by people and agents.

In the future, the real gap between companies will lie in who can embed AI into business flows, turn experience into organizational assets, and form an organizational system that can continuously operate and evolve.

AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?
AI 已经进入企业,为什么组织还没有进化?

This is an original article from Leiphone.com; unauthorized reproduction is prohibited. See the reprint notice for details.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by 雷锋网
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗