Author: Shuhe
Source: Zhenghe Island (ID: zhenghedao)
Tencent has made another big move.
From October 29th to 30th, the 2026 Tencent Global Digital Ecosystem Conference was held, sparking heated discussions around the world. This conference, with the theme of "Group Intelligence Collaboration Digital Flywheel: Stimulating Productivity in All Scenarios", focused on the in-depth application of AI and Agent in thousands of industries.
As AI becomes more and more powerful, how companies transform, improve organizational efficiency, and release people's maximum potential has become the focus of attention for companies and individuals.
It is worth mentioning that Tencent set up a special session for prosumers (professionals) for the first time at the conference, in which Tencent's HR, administration, legal affairs, finance and other functional lines shared their AI practices in public for the first time. In just one year, Tencent internally achieved deliverable, verifiable, and evolvable organizational-level AI productivity.
Facing the wave of AI, how can a large enterprise evolve its organization and transform into an AI-native organization? Tencent’s internal AI practices provide enterprises with reference and reusable organizational evolution samples.
2026, an organizational evolution
Every productivity revolution will bring about a particularly large organizational revolution.
In the agricultural era, it was land, the basic organizational unit was the family, self-sufficiency, and the social organization method was the expansion of blood and geographical relationships. In the industrial era, it was capital and machines, so there were positions, processes and bureaucratic management. In the information age, there were professional skills, better information matching, SaaS and knowledge base management, so knowledge-based enterprises emerged.
In the era of intelligence, AI is becoming more and more powerful. It can exhaust the vast amount of public knowledge of mankind, complete calculations, deductions, sorting, output and other standard actions, write articles, write code, make designs, generate short plays...it is proficient in eighteen kinds of martial arts.
More and more companies and individuals are beginning to take action to explore how to use AI for their own use, using it to help them reduce costs, increase efficiency, and innovate business. However, they find that it is not easy to use it.
One of the most real dilemmas in the implementation of AI is that individual efficiency has doubled, but organizational efficiency has not increased significantly. McKinsey's survey data is even more disturbing. Nearly 90% of companies around the world are engaged in AI transformation, only 25% have achieved tangible results, and 75% have been internally consumed in pilot projects for a long time.
Li Ning, a professor at the School of Economics and Management at Tsinghua University and an organizational research scholar, has concluded that the AI transformation of most enterprises is a false proposition.
Where does enterprise AI transformation get stuck? The biggest bottleneck in the implementation of AI is not the technology itself, but the organization.
The first is organizational inertia. Professor Li Ning once said a very vivid metaphor. He compared the AI transformation to renovating an old house full of employees. The company has a complete division of labor, and everyone performs their duties step by step. After the introduction of AI, the original business continues, and "renovation" will affect the business progress.
The second is the resistance of employees. AI is becoming more and more powerful, and some entry-level jobs have been taken over. Employees are afraid of being replaced by AI, so they are psychologically resistant, unwilling and afraid to use AI. Some companies will also experience cognitive dislocation and use AI as a "layoff tool". The original intention is to improve efficiency, but once problems arise, AI cannot solve the problem.
Third, traditional organizations are people-centered and designed around "people." The use of AI by enterprises is simply equivalent to "old business + tools" and does not achieve true "organization and business reconstruction." As AI gradually enters business processes, the original job boundaries, collaboration methods, and decision-making mechanisms of enterprises have all changed.
After a new technological revolution, how will humans divide labor and cooperate? What will the organization of the future look like? When more and more jobs are replaced by AI, where will the future value of people be reflected? This is an issue of the times that businesses and individuals are concerned about.
In the past year, Tencent has embarked on an internal organizational evolution. The four major functional departments of human resources, administration, finance, and legal affairs have boldly innovated and started from the pain points of front-line business to solve the difficulties of enterprise AI transformation. Their AI practices will provide enterprises with examples that can be learned and thought about. Let’s explain them one by one below.
HR｜From personal amplification to organizational evolution
Turning the AI-native productivity flywheel
In the traditional HR (human resources) department, a large number of manpower is tied up in repetitive tasks such as resume screening, attendance accounting, and salary tabulation.
The deeper contradiction is hidden in the data layer. Personnel data alone is scattered in multiple systems, forming a natural "data barrier." Due to different ownership definition standards, security concerns and other factors, data is difficult to circulate, integrate and share across domains, and can only rely on manual aggregation in Excel, resulting in a lot of invisible work.
When HR's time for core work is severely squeezed, there is very little time left for talent inventory, organizational diagnosis, and talent echelon building.
In order to change the status quo, many HR teams will put forward demands to the technical department, such as establishing a recruitment model, but are limited by work processes and operating costs. A demand often takes more than three months from review, development, and testing. However, it is difficult to agree with everyone, and the models developed may not necessarily be what front-line departments need. This is also the most painful and helpless part of the HR team.
The emergence of AI has given Tencent HR team a solution.
Within Tencent, the HR team was the first to explore AI practice and is also one of the most mature departments in AI practice. They launched a platform - HRClaw, and in just six months, they have accumulated more than 3,000 applications, with an average daily call volume of more than 14,000. It is worth mentioning that these applications turned out to be handmade by a group of non-technical business and HR teams.
How? It sounds simple. With the help of tools, we can reshape the workflow and cultivate the "super organization" soil of AI.
Everyone with specific needs can put forward their needs in WorkBuddy, generate previews, access data, build a real-time workbench in 30 minutes, and then share it directly with those who need to use it. The use and interaction of different permissions and roles are different.
After running it, the effect is very significant. The technical development workflow has undergone tremendous changes. It was originally in the form of "proposing requirements-scheduling-delivery and acceptance", at least 4-5 flows to the process, which took more than 3 months. Now the HR team can do it themselves according to the needs, and can complete development and go online in as fast as two hours.
The bigger change is on the R&D side. Its data dashboard requests have dropped sharply, but data service call requests have skyrocketed. This is no longer just a flash of inspiration from a "super individual" who knows AI very well, but a "super organization" that allows more people to get started with AI without any barriers.
It all starts with data governance.
HR is a naturally data-sensitive department, and there are two major difficulties: On the one hand, the data "does not match up": recruitment is based on progress, training notices are sent, incentives are collected, activities are reviewed, and the fields of the same data in different tables are different;
On the other hand, data is "dare not to be given": data is highly private, and HR data involves a large amount of sensitive information such as salary and performance. However, the data cannot be controlled too tightly, and how to loosen permissions is a difficult problem.
To this end, the HR team has built a five-layer authority control logic, and based on this, a digital intelligence expert has been launched.
The so-called HR digital intelligence experts are built based on the expert team capabilities of the WorkBuddy platform and are oriented to the company's intranet.
When employees call WorkBuddy, they have their own identity recognition: for the same problem, people with different management scopes obtain different data and obtain different analysis results; similarly, the data placed in the background also comes with its own permissions. When the model and the agent access the same data wide table, the public network large model can only access low-sensitive data, the customized intranet large model can access highly sensitive data, but the core data cannot be read by AI.
When data is read and employees access it, they all have their own permissions, so more things can be done. An HR digital intelligence expert that needs to be called from WorkBuddy is upgraded to a systematic solution for corporate intranets, including data governance, application deployment, release and sharing.
Based on the 5-layer authority data platform, HR has also built 3 pipeline workflows, specifically:
Generation end: Users put forward requirements in WorkBuddy, and page-deliver calls internal data and knowledge and generates previews;
Publishing side: triggered by HRClaw and deployed to the OA K8 production environment;
Access side: Users obtain deployment-free and shareable links through unified Taihu authentication, identity resolution, and permission verification.
In the past, "employees asked data and asked questions" was a one-way output, but now it has become a cyclical and sustainable productivity of "employees build a workbench - share it with users - user data feedback accumulates - further optimize the workbench".
From data permissions to no threshold for release. HRClaw has begun to realize a positive cycle of self-growth.
On the HRclaw platform, everyone can create their own applications. New applications will attract more users and accumulate more data, and more data will give everyone more possibilities to develop scenarios.
For example, professionalism is like Tencent's recruitment experts, which allow AI capabilities to cover the entire recruitment process. They can search resumes, come up with interview questions, and issue evaluation reports 24 hours a day, 7 days a week. Their professionalism exceeds that of most junior interviewers; verticality is like newbie training. All newcomers can directly see what they need to do by using "Newcomer Goose Village" and start training classes with one click; see who should be the closest partner in the "work chain" and start corporate and micro-communications with one click.
Administration｜Knowledge accumulation thickens organizational assets
From "details manager" to system designer
Administration has always been the "big housekeeper" in the enterprise. Daily affairs are complicated. During the day, they spend time on resource management and inventory, organizing company meetings, planning welfare activities, etc.
What’s even more worrying is that the data and information are scattered in more than a dozen forms, dozens of group materials, and hundreds of emails. Each query, calculation, and analysis will consume a lot of time.
They are also "firefighters". They come to them for work station repairs, activity innovations and employee atmosphere. It is normal to be so busy that they don't even touch the floor.
In the past year, the technological changes brought about by AI have drastically changed the way administrators work.
At the conference, Zhao Ge, administrative director of Tencent Group, mentioned that last year, administrative staff only used AI to do some document sorting and material production, but today, the resident intelligence on the company's servers can detect data at all times, detect problems in advance, generate work orders, provide feedback, and give long-term optimization suggestions.
"AI not only makes document sorting faster and problem discovery more timely, it also frees up our energy to improve services, turn ideas into experiences, and make good products available to more people," Zhao Ge said.
Zhao Ge, Executive Director of Tencent Group
In the face of AI changes, the evolution path of Tencent's administrative team is to use a comprehensive management platform, work station management, administrative budget, assets and equipment and other resource status, which can be clearly seen on the dashboard, basically covering most work scenarios.
The greatest value of this platform is to change administrative services from "passive response" to "predictable and early warning".
On the billboard, all warning matters are marked in different colors such as red, yellow, and blue according to their degree of urgency. When encountering an emergency mark, click on the contact person to be directly linked to the corporate WeChat for one-to-one communication and efficient problem solving, forming a closed loop of workflow.
For example, in the case of gift inventory, administrative staff can control the entire process of purchasing, warehousing, use and delivery through the Kanban. When the inventory is lower than the set threshold, an early warning is automatically triggered and one-on-one reminders are sent through corporate WeChat.
At the scene, Zhang Wenli, assistant to the vice president of Tencent Cloud and senior lecturer of Tencent Academy, told a moving story. She said that when she first joined Tencent, she happened to see a concept in a bookstore that still shocked her - nine-stage secretarial work.
Zhang Wenli gave an example of a meeting. The first-stage secretary is responsible for issuing notices, the second stage is for implementation, the third stage is for re-inspection, and all the way to the ninth stage, the entire process is standardized and passed down. In the past, such experience required repeated polishing, but now, by sending an instruction to WorkBuddy in the corporate WeChat, the ninth-stage secretarial work can be completed in one step.
Zhang Wenli, Assistant Vice President of Tencent Cloud, Senior Lecturer at Tencent Academy
This is why AI has brought about a jump in value and expanded the boundaries of human capabilities. But AI is not just limited to office scenes, even inconspicuous high-frequency scenes such as ordering food have also been reconstructed.
Ordering food is a seemingly simple scene, but it is very time-consuming. In the past, administrative staff needed to check the ordering location, environment, distance, cost, etc. on major apps. Now, just say a word to WorkBuddy and get a list in the form of a web page for reference and selection in 2 minutes. During the 2 minutes of waiting, the administrative staff can also use WorkBuddy to prepare a monthly administrative report.
In terms of workstation management, under the traditional model, administrative staff need to spend a lot of time coordinating workstations and debugging equipment. Now Tencent administrators can intuitively observe the utilization rate of workstations, ratio health, workstation costs, etc. through the comprehensive dashboard. With accurate data statistics, more reasonable management suggestions can be given for administrative resource allocation and administrative expense planning.
When tasks, progress, and responsibilities can all be seen, tracked, and reminded on the platform, administrators no longer need to play the role of "firefighters", but evolve into designers of the system - setting permissions, boundaries, and audits. This is an upgrading of the capability structure.
But no matter how smart the system is, it cannot "run completely on its own." Where it runs, how fast it runs, and what kind of atmosphere and care employees need, it all depends on people. Tencent executives make good use of AI to liberate people from repetitive work, while they focus on decision-making, overall control, and humanistic care. This new paradigm of complementation and co-creation between humans and AI is particularly evident in the reshaping of employee experience.
For example, Tencent's administrative secretaries are good at planning and innovation. They use WorkBuddy to create a healthy and uplifting sports atmosphere in the department; they use WorkBuddy to create various festival mini-games to enhance employee participation with personalized interactions; they use team-building manuals to automatically capture weather forecast skills and add the function of autonomously selecting roommates so that each participant can intuitively understand the key points of the itinerary. AI is infinitely expanding the boundaries of human capabilities.
Zhao Ge said: "AI stimulates new administrative productivity, which ultimately comes down to people's feelings: making employees less troublesome, making the team more calm, and making the organization run better. "
AI is responsible for "improving efficiency", while people are responsible for coordination, control and "carefulness". The more advanced the technology is, the more people are needed to guard the temperature of the organization.
Finance｜Human-machine collaboration, organizational growth
Financial people and intelligent agents "learn by doing"
In a company, the financial officer is often called the "Mr. Accountant" and is responsible for reimbursement review, invoice verification, month-end settlement and other tasks. This type of affairs has strong tidal characteristics. For example, the peak period of business signing is usually concentrated at the end of the month, and shortage of manpower has become the norm.
In addition, finance is highly professional, and a large amount of professional knowledge is implicit and locked in the minds of senior employees. When junior employees encounter complex problems, they can only wait for guidance from experienced colleagues; and sorting out these implicit experiences into rules is itself a long and complicated process.
At this ecological conference, Zhang Ying, director of financial management of Tencent Cloud, said: "AI allows finance to move from 'bookkeeping' to 'helping enterprises make better decisions.' "
AI helps Tencent Finance achieve a new work paradigm of "digital self-sufficiency". Tencent's finance team builds a unified data warehouse, uses AI's vibe coding capabilities to directly write code, customize and develop BI dashboards, and supports real-time query data, decision simulation, business analysis and other scenarios, allowing management to obtain business results earlier and improve the speed and quality of decision-making.
Zhang Ying, Director of Financial Management of Tencent Cloud
AI has improved work efficiency. The processing time of many tasks has shown a stepwise transition from "days" to "hours", and then from "hours" to "minutes", which has significantly improved efficiency.
In daily business analysis scenarios, financial personnel have made many innovations. In addition to customized dashboards, they have also accumulated analytical experience into reusable business analysis skills. With WorkBuddy, the team can initiate real-time inquiries on business indicators, quickly locate the root causes of fluctuations and quantify their impact, significantly improving analysis efficiency.
At the same time, the team introduced AI tools to build a product revenue forecast model. By dynamically adjusting core factors such as volume and price, it simulated revenue changes under different combinations in real time, providing immediate and quantitative support for product pricing and supply decisions.
During the exchange, Tencent's financial staff demonstrated their own AI practices. Through natural language and a simple command, corresponding data can be automatically generated, and finally decisions are made by humans.
To give an example of daily payment collection management, open WorkBuddy, call the receivables analysis skill, and enter "Generate September receivables data" in the dialog box. The system will automatically generate a payment receivable monitoring dashboard within seconds. The dashboard presents multiple core indicators such as customer income, payment cycle, and credit limit in a centralized manner. The fluctuation trend is clearly visible, helping the team to discover potential payment risks in a timely manner and providing data support for subsequent collection actions.
In the corporate WeChat, financial personnel can check the amounts to be written off through the "Writing-off Assistant" robot. After manual confirmation, the "Writing-off Assistant" automatically matches each amount and provides a variety of write-off plans. The processing time of a single transaction is reduced from 5 minutes to 1 minute.
In response to the shortage of manpower during peak business periods, Tencent's financial staff encouraged everyone to use more, try more, and introduce AI to improve quality and efficiency in just four steps on the premise of data compliance: first sorting out the workflow of each module, then identifying repetitive, inefficient, and error-prone links, then matching appropriate AI tools, and finally promoting replication through training and knowledge bases.
Financial work is highly dependent on personal experience, especially for non-standardized contracts. 18 types of risks and 8 items must be reviewed one by one, and the judgment standards are different. Different reviewers may look at the same clause, and the conclusions may be completely opposite.
The solution of Tencent's financial team is to "skill risk identification" - to precipitate the review rules into a set of standards that can be executed by AI. Financial personnel define the risk judgment caliber and update and iterate the risk identification skills; when approving contracts, WorkBuddy calls the contract review skills, which can automatically generate a complete, multi-dimensional risk analysis report, assist finance in contract review, and finally confirm the approval manually.
Every conclusion has its original source, is automatically traced, and is traceable. If the rules are manually changed, historical contracts can be run back to verify the effect. This is to turn "relying on experience" into "relying on standards."
After running in and trying, Tencent’s financial staff have formed a human-machine collaboration model of “90% AI execution + 10% manual confirmation”, where AI is responsible for execution and humans are responsible for final confirmation.
In order to transform personal experience into organizational capabilities, Tencent's financial personnel have built and accumulated a team knowledge base, turning financial expertise into an organizational asset that can be retrieved and called by AI. It has been implemented into an intelligent question and answer system, thinking about answers based on the content of the knowledge base, and creating a "financial know-how/little assistant" that can directly ask questions.
AI Q&A has been online for several months and has been used thousands of times, covering 300 people/month. The average response time for consultations is less than 10 seconds. The experience has been accumulated into an asset that the company can retain and AI can use.
In traditional organizations, what is most easily overlooked is the growth of people. From interns to senior employees, it is driven by people's growth and experience to improve the organization's capabilities. AI breaks this path, and the skills accumulated by key employees make capabilities replicable and shareable. At the same time, individuals are liberated from repetitive tasks and can devote more energy to new tasks with more creativity and higher value, expanding the space for personal growth.
Tencent AI practice, some inspirations
Entering 2026, AI's impact on organizations is not just at the tool level. It subverts the hierarchical organizational model that has been based on the Industrial Revolution for hundreds of years. As AI is used more widely, there are more "super individuals" in enterprises, but for corporate productivity, many "super individuals" will not automatically combine into "super organizations."
Chinese companies should rethink organizational management and talent strategies in the AI era, and Tencent’s internal AI practices provide a reference and reusable organizational evolution path for organizational AI upgrades.
1. Cultivate super individuals. Within Tencent, outstanding employees use AI extensively. Some people develop comprehensive management platforms, some train skills, and some invent human-machine collaborative development models, reshaping workflows and exceeding efficiency benchmarks.
2. Deeply embedded in scenarios. In enterprises, AI cannot just improve individual efficiency. Starting from front-line business pain points, AI should be actively used to develop new needs, solve problems, redefine positions, re-engineer business processes, and let AI be deeply embedded in business closed loops.
3. The management objects have changed. In the past, the traditional organizational model was to manage people, and the hierarchical management system relied on the managers’ personal abilities. When AI is deeply embedded in the workflow, in addition to managing people, AI and workflow should also be managed.
4. AI does not replace humans, the real value is creation. In the past, humans were “users” of tools, but in the future, humans and AI will have a deeply collaborative relationship. AI is best at repetitive, structured tasks, while humans’ unique judgment, creativity, self-drive and other abilities are rapidly appreciating.
5. Build a super organization. Tencent not only has many super individuals, but also a group of dedicated agents, shared knowledge bases, standardized workflows, etc., which together form a human-machine symbiotic network, moving towards a new organization driven by intelligence.
The future has arrived. In the face of the roaring technological revolution, how to adapt production relations to the leap in productivity as soon as possible has become a realistic proposition that every entrepreneur must answer. Change will not stop because of people's hesitation. Only by seeing the trend clearly, actively embracing it, and accelerating change can enterprises weather the storm and win the future.
Typesetting | Phase and
Reviewer | Xingchen Editor-in-Chief | Xia Kun