From Digital Employees to Knowledge Productivity

Artificial intelligence (AI) is entering a new phase in its relationship with business. In the first stage, companies naturally focused on tools, models, computing power, and efficiency. These were the most visible aspects of the technology, and they were also the easiest for managers to understand. Yet, as with earlier waves of technological change, the more important issue is not the tool itself but whether the tool can be absorbed into the company’s way of working, learning, serving customers, and creating value.

This distinction is critical. Many companies today have already begun to use AI, but not all of them are becoming more capable as organizations. Employees may use AI to draft documents, summarize meetings, generate reports, support customer service, or conduct preliminary analysis. These applications are useful, and in some cases the efficiency gains are immediate. However, most of them remain at the level of individual productivity. If a company stops there, it may overestimate what AI can do and then become disappointed when the broader impact fails to materialize.

AI Must Enter the Core of the Organization

The next stage of AI application will be judged by a different standard. The question will not be how many people in a company are using AI, or how many tools have been purchased. Instead, it will be whether AI can enter the company’s business scenarios, knowledge systems, and value chain, and over time become a sustainable source of organizational productivity. In other words, the real measure of AI adoption is not usage. Rather, it is whether AI changes the way the enterprise creates and renews its capabilities.

This is where the notion of the “digital employee” becomes useful. AI agents are gradually moving from concept to practice. They are no longer simply helping employees write a paragraph, retrieve information, or complete a simple query. Increasingly, they can take on specific tasks inside business processes and become extensions of human capability. Their significance lies not in how advanced the technology appears but in whether people can lead, train, supervise, and deploy these digital employees in real business environments.

From Local Efficiency to Organizational Productivity

The first effect of digital employees is process redesign. Many activities that are repetitive, standardized, and attention-consuming can now be handled in part by AI: collecting information, applying rules, building tables, preparing meeting summaries, responding to customers, and generating operating materials. These tasks may look ordinary, but they have long consumed a large amount of organizational energy. When AI begins to take them on, people can spend more time on judgment, communication, innovation, and customer relationships. For companies, this is not only a productivity improvement but a reallocation of human capability.

At the same time, process redesign should not be confused with automation. If AI is simply inserted into an existing process, the company may become faster without becoming fundamentally better. The more important opportunity is to ask whether the process itself should be redesigned. In retail, for example, leading companies are not only using AI agents to help employees. They are beginning to apply them to customer decision-making, supplier negotiation, inventory management, advertising, and ecosystem development. In manufacturing, the pace may be slower because production must interact with the physical world. But in factories that already possess strong digital and automation foundations, AI agents can become assistants to plant managers and operating managers, helping them manage core production activities.

The broader implication is that AI’s impact on the value chain will not stop at internal efficiency. Over time, the interactions among functions inside a company, and between the company and its customers, suppliers, channels, and ecosystem partners, may increasingly involve coordination among agents. Consumers may also gain more power in the process. In the past, their choices were shaped largely by brands, channels, and marketing systems. In the future, agents may help consumers clarify intentions, articulate needs, and complete transactions. As this happens, retailers, manufacturers, and service providers will all have to reconsider their roles.

However, a company is not a collection of isolated tasks. It is a system of interdependent activities. If one part of the process becomes more efficient while the upstream and downstream links are not adjusted, the value created will remain limited. The next stage of AI application will therefore depend on whether agents can connect with existing software, cross-functional workflows, and internal collaboration mechanisms. Without this connection, AI will remain effective for individuals but ineffective for the organization.

Knowledge Is the Foundation of AI Capability

The deeper foundation for this connection is knowledge accumulation. Whether an AI agent can perform useful work depends on whether the company’s knowledge has been captured, expressed, and circulated. Explicit processes, standards, and rules are important, but they are not enough. Much of a company’s most valuable knowledge is tacit. It resides in the judgment of frontline employees, veteran technicians, outstanding store managers, experienced salespeople, and senior managers. This knowledge is often difficult to write down in formal documents, but it is frequently what gives the company its distinctive operating capability.

Consider a maternal nutrition adviser who sells infant formula well. Her real capability does not necessarily begin with recommending a product. It may begin with building trust with the mother, understanding her situation, and identifying the problem she is trying to solve. Or consider an experienced operator in a beverage chain business who, during a store-opening inspection, can solve a food-safety issue without forcing the store to rebuild. These may appear to be small examples. In fact, they reveal a larger point: much of what companies call experience is accumulated judgment formed through repeated encounters with real situations.

This is why knowledge management in the AI age is different from knowledge management in the past. Previously, many companies found knowledge management difficult because it required employees to organize, upload, and maintain information outside their normal work. Today, if communication, meetings, documents, and business collaboration take place on integrated platforms, AI can help extract knowledge as part of the work itself. Knowledge accumulation no longer needs to be a separate administrative exercise. It can become part of daily operations. Only then can digital employees gradually understand the company, the business, and the customer.

After knowledge has been accumulated, it must circulate. Companies generate enormous amounts of information through sales meetings, store meetings, operating reviews, and customer interactions. Human managers cannot read all of this material line by line, nor can they continuously identify patterns across every record. Large models can help extract operating signals and produce insights about market activities, expense allocation, regional performance, and execution issues. AI’s value is not merely that it records information. It is that it can turn dispersed information into judgment that management can use.

This, however, brings companies into the more difficult area of data governance. Traditional SaaS and IT systems were largely designed for human users, with interfaces built around what people need to see and operate. AI often requires more direct and complete access to enterprise data. This creates new questions. How should the data warehouse be opened? Which data can be processed by cloud-based models, and which data must remain with local models? How should access rights, audits, and accountability be managed? These questions are not technical details alone. They are part of the governance architecture that will determine whether AI remains experimental or becomes a true enterprise capability.

Ping An Life offers a useful example of how this progression can unfold. Its AI journey moved from intelligent service to digitalized operations, then to broader AI applications and service upgrades. In the early stage, intelligent service agents could handle large volumes of calls. Later, through digitalized business management, operations, and administration, the company built a digital brain that connected operating indicators, process indicators, and employee responsibilities. Employees could see what was happening in their areas of work and anticipate what might happen next. AI then began to support sales, customer insight, opportunity recommendation, communication content generation, internal administration, compliance, consumer protection, and coding.

More importantly, AI began to make service more direct and integrated. When a customer expresses a need such as claiming insurance benefits, the system can identify the relevant policies and rights under the customer’s name, calculate the benefits available, and complete the operation after confirmation. In global emergency assistance, AI can interpret a customer call or SOS message, generate a risk code, match a rescue plan, and mobilize rescue resources. These cases show that AI becomes a true enterprise capability only when it enters customer service and the business loop. At that point, it is no longer simply an internal tool. It becomes part of how the company delivers value.

This does not mean that digital employees should be understood as substitutes for people. A new division of labor is required. Digital employees may be better at rule-based judgment, information organization, standard responses, and repeated execution. People remain essential for communication, listening, empathy, complex judgment, relationship building, and sensitivity to context. In some management situations, AI may generate feedback that is accurate but too direct, missing the nuance required in human communication. AI can help companies do things correctly. People must still help them do things appropriately.

This is one of the important paradoxes of the AI age. As machines become more capable, companies must become clearer about what people are uniquely able to do. The answer is not that people should retreat from work. Rather, they must move to higher-value parts of work: judgment, interpretation, relationship building, ethical choice, and responsibility for outcomes. Companies that understand this will not treat AI as a simple labor replacement. They will use it to redesign the relationship between people and work.

In this process, people who know how to use AI well should become central to organizational development. Companies need to identify employees who understand the business, are willing to understand AI, and possess critical thinking and judgment. Several types of roles will become more important: people who can use AI to achieve the output of much larger teams, people who can take responsibility for business outcomes, and AI coaches who can cultivate and activate such people. Over time, computing power, tools, and agent quotas may themselves become new forms of incentive resources.

For business leaders, the idea of a management avatar will also become more relevant. A manager’s weekly reports, meeting comments, decisions, and written feedback can help AI understand his or her management style and provide support in review, judgment, and communication. Yet this does not mean the manager can withdraw from management. Rather, the more powerful such an avatar becomes, the more important it is for the manager to have clear values, judgment standards, and organizational requirements. AI can amplify managerial influence, but it cannot replace the leader’s underlying judgment.

The way companies measure the value of work will also need to change. When employees use AI to complete validation, analysis, and prototyping that previously required more people or more time, companies cannot continue to evaluate contribution only by hours worked, documents produced, or process milestones completed. Strategic researchers, business analysts, and product managers should not be dismissed simply because AI can generate materials. Their value is shifting from producing documents to asking the right questions, testing hypotheses, and forming workable solutions.

This change will alter the path from idea to validation. In the past, any product or strategy idea often required design drafts, product requirement documents, animations, and multiple rounds of meetings before serious discussion could begin. Today, professionals can use AI to create working demos more quickly, after which engineering teams can take over for large-scale deployment and professional development. Even in hardware, some solutions can first be tested in virtual environments before entering development. The point is not that non-specialists should replace engineers. Rather, professional collaboration can move earlier into a higher-quality validation stage.

From Scattered AI Use to Enterprise Productivity

The key to AI implementation is not how much software a company buys. It is whether the company can connect digital employees, knowledge accumulation, process redesign, data governance, and business value. If AI is used only in scattered ways by individual employees, the company may gain individual efficiency. If AI is embedded into core processes, continuously accumulates knowledge, replicates experience, and closes the loop with business outcomes, the company may begin to achieve organizational productivity.

Looking forward, AI agents will increasingly enter not only the enterprise itself, but also the relationships between companies and their customers, suppliers, channels, and ecosystem partners. Companies will need to ask not only which tasks AI can perform, but which knowledge it can absorb, which processes it can reshape, which employees can become AI pioneers, and which services can become more precise and more human because of AI.

Ultimately, AI is not the endpoint of enterprise transformation. It is a new means through which companies can rebuild knowledge, processes, and value creation. Only when AI moves from tool to digital employee, from digital employee to carrier of organizational knowledge, and from knowledge carrier into the value chain and customer service loop, will it become a real new source of productivity. This process is still unfolding, and how companies manage it will help determine their position in the next stage of competition.

This article was originally published by Gao Feng Advisory Company in June 2026.