Five AI implementation narratives reveal a "corporate confession session"

The French social psychologist Gustave Le Bon wrote in his book The Crowd: A Study of the Popular Mind: The crowd doesn't need the truth; it only needs a definite answer. This statement finds a highly relevant interpretation in the era of AI.

When people first encounter generative AI, their most instinctive reaction is to ask it a question and then wait for an answer. The value of AI seems to lie in how quickly and accurately it can provide a definite output. When companies purchase AI, the same narrative is presented in their PPTs, such as how much manpower can be saved by intelligent customer service and how much efficiency can be improved by AI-assisted programming.

Recently, market research firm IDC released a set of figures: 81% of organizations are deploying, piloting, or planning to adopt AI PCs in the near future, and 61% of enterprises have directly integrated AI into their work processes. The top three factors driving enterprises to invest in AI PCs are productivity improvement (59%), innovation and competitive differentiation (39%), and security benefits (35%).

However, the willingness to deploy does not equal the deployment effect.

The report also shows that the cognitive gap between CEOs and frontline employees is widening. 32% believe that employees have limited understanding of AI functions, and 31% are concerned about the uncertainty of regulatory and data compliance.

DoNews has contacted professionals from various industries, including banking, insurance, auditing, technology, and real estate. In the process of their active or passive embrace of AI, a clear yet complex picture emerges: some people regard AI as a "universal key" for cost reduction and efficiency improvement, while others become a "buffer layer" between algorithms and manual work under performance indicators and compliance pressure.

01

The most primitive state of AI implementation is not failure but falsehood. Some enterprises regard AI as a "high-tech veneer" that can be casually pasted on, with a huge gap between their cognition and actions.

Ziming used to work in the auditing industry. He shared with the author the attitude of his former company towards AI - from a "lobster-style" mobilization to hesitation in actual implementation.

According to him, the boss once issued an order in a company group of more than 200 people, asking everyone to "put away the lobsters." However, at that time, few people in the R & D department really knew how to use the relevant AI tools, and no one had evaluated the data security risks and the potential Token fees for API calls in advance.

The senior management verbally expressed their willingness to "embrace AI," but in terms of actual actions, the resource support was almost zero. Ziming admitted that he once tried to build an enterprise knowledge base with "Coze" and had successfully completed the self - test, but finally got stuck in the docking of the enterprise WeChat interface. "Because there was no budget at all." Some colleagues also deployed some AI applications on Feishu on their own, but the overall proportion was not high.

What is most confusing is the boss's inconsistent attitude towards AI. "The company has broad prospects and space for AI applications, and the boss often participates in various AI - themed conferences and activities. But once after coming back, he told us that manual work was still better and that AI could never replace our work."

Oscillating between technological worship and implementation resistance is a true portrayal of some enterprises' embrace of AI. The same sense of disconnection also appears in another form of "forced embrace."

DoNews contacted a programmer from a listed company. According to him, the company has been promoting "AI for all" since January this year, requiring employees to use "designated AI tools" to generate code and promoting the transformation of all employees into full - stack engineers. However, there is an obvious gap between the actual effect and the original intention.

For example, the leadership evaluates the time consumption of AI development and traditional development methods every day, resulting in a sharp increase in the number of meetings and overtime work. "This somewhat deviates from the original intention of AI to improve efficiency." He believes that the company's leadership is overly optimistic about AI's ability to solve most development problems, while frontline employees are exhausted by the daily evaluation meetings and overtime work.

The executives expect AI to improve productivity, but employees only feel an increase in workload. Some employees even don't know how to achieve the leadership's expectations. Inside the company, employees are also privately discussing another sensitive topic: since they are all programmers, the most suitable way to use AI for themselves is the best. Why do they have to be forced to use the designated AI generation method? Excessive meetings, passive overtime, and concerns about data usage are weakening the positive experience that AI should bring.

The employee helplessly raised a question: Is AI used to make employees' work more flexible or to increase the company's production capacity? Does AI help workers or ultimately replace them? These are not just personal confusions but the common anxieties of countless front - line technicians after AI enters the workplace.

02

When enterprises start to take AI seriously, invest resources, and build systems, the transformation moves from "slogans" to the "organizational level." But at this stage, the improvement of efficiency often brings pain at first.

Kaka works for a large real - estate company in China. He introduced to the author that after the launch of DeepSeek in early 2024, the company officially started its AI - enabled journey.

The first stage is learning. "The company will issue meeting notices and provide video materials, requiring us to learn about AI." In the second stage, the professional team will integrate AI more deeply with the internal system and independently develop multiple tools to improve enterprise efficiency.

"AI can indeed extract useful information and analyze data, making the positions more professional. The other day in a meeting, the leader was trying to teach us how to 'feed the lobsters,' but the time was too short, and our absorption was limited." Kaka explained that the leader's main intention was to raise employees' awareness of AI and help them actively use AI in their daily work without resistance.

After the improvement of efficiency, enterprises often re - allocate tasks, and the most obvious change is reflected in the organizational structure. It is understood that the company launched a round of organizational structure adjustment this year. Previously, there were three levels: grass - roots positions, manager positions, and director positions. Now, the grass - roots positions and manager positions have been merged into one.

The change is happening rapidly, and employees' mental states are also diverging. Kaka observed two trends: some people are very interested in AI, actively learn, and can move up on their own; others are still using methods from the past few decades or even the last century to do work in the new century and are unaware that the era is leaving them behind. These people can only be guided by the enterprise.

From the initial learning to the subsequent increase in workload and then to the merger of positions, the rhythm of work content change is very fast. "The senior management can't care too much about the long - term, and employees are indeed quite confused," Kaka said.

03

The way to get out of the pain is not to slow down but to find the right focus. When enterprises start to precisely integrate AI into high - value, standardized business processes and build confidence through quantifiable efficiency improvement, AI truly transforms from a "concept" into a "tool."

Xinmei Mutual Life Insurance (hereinafter referred to as "Xinmei Insurance") in the insurance industry provides a mature model. The relevant person - in - charge told the author that the company's AI layout has moved from the initial trial stage to the large - scale application and efficiency - improvement stage, and AI technology is profoundly changing its core business processes.

As a financial institution under strict supervision, Xinmei Insurance adopts a differentiated strategy in computing power deployment: it uses cloud services for non - sensitive business and private servers for sensitive business to ensure customer privacy and core data security.

In terms of strategy, Xinmei Insurance adopts a light - asset model, with all computing power resources coming from the cloud and no self - built computing facilities. The cost is mainly focused on technology research and development and talent cultivation. In terms of measuring the input - output ratio, the company focuses on three dimensions: operational efficiency improvement, cost control, and risk control, and tracks core indicators such as underwriting and claims settlement timeliness and manpower cost savings.

It is worth noting that they do not set up the so - called "AI KPIs" in the traditional sense but deeply integrate goals such as technology implementation, efficiency improvement, and risk control with the business. "Employees generally have a positive attitude towards AI. Many people set more challenging goals through OKR and actively propose suggestions for strengthening AI training," said the person - in - charge of Xinmei Mutual Life Insurance.

This practical strategy has brought tangible changes. Take the pre - underwriting of health insurance as an example. The AI intelligent pre - underwriting assistant can automatically identify medical document information and conduct analysis and reasoning based on underwriting knowledge, significantly reducing the pre - underwriting time from the traditional 48 hours to less than 1 hour.

04

When the tool is powerful enough to replace standardized labor, the value of human beings no longer lies in the execution level but in the ability to master the tool and make complex judgments.

The practice of China Merchants Bank (hereinafter referred to as "CMB") provides an example for this proposition. As a well - recognized "leader in technology - driven finance" in the industry, CMB has made substantial investments in AI.

Data shows that in June 2025, CMB officially announced its "AI First" strategy. In this year, there was an explosive growth in CMB's AI technology and applications. The daily average Token throughput in 2025 increased by 10.1 times compared with 2024. Throughout the year, 183 specialized models in the financial vertical field and 856 business scenario applications were implemented, and AI replaced 15.56 million hours of manual work.

The author contacted Shu Lan, a corporate credit manager from a CMB branch in a certain city. Her description shows a calm and practical attitude of bank employees in the AI era. Shu Lan said that in recent years, CMB has allocated a significant amount of funds (about one to two billion yuan) from its revenue and profit to the technology - driven finance sector every year. "It is the strongest in the industry, no one else can compare." This huge investment has made AI permeate every aspect of marketing, customer service, and even official document writing like air.

According to Shu Lan, in the traditional model, when customers consult the company's loan business through mobile banking or the official website, they are often limited by the busyness of manual customer service and professional barriers. Now, the intervention of AI is completely breaking this bottleneck. The response speed of AI is almost instantaneous, and more importantly, there is a leap in professionalism.

"In the past, when facing complex products, manual customer service often had to transfer calls or consult materials. But now, the dedicated AI system has internalized a vast amount of product details and compliance requirements. It can not only accurately answer customers' questions but also present clear logic and standardized language, which is far more reliable than manual rote memorization," Shu Lan said.

In addition to customer consultation, in the external marketing process, CMB's self - developed AI system will capture customers' dynamics and potential needs through big data and actively generate marketing leads and push them to customer managers. "For example, if the system detects abnormal fluctuations in a company's capital flow or signs of expansion, AI will immediately push business opportunities for me to follow up precisely. It can even automatically generate visit logs and communication summaries and directly upload them to the system."

Working with regulatory authorities is also one of Shu Lan's daily tasks. She said that in the past, it took one or two days and a lot of effort to draft an official document report. Now, she only needs to "throw" the outline and format requirements to AI, and a standardized draft can be generated in a few minutes, which only needs minor adjustments. The saved time allows her to focus more on in - depth analysis of customers and risk control.

What interests the author more is that in CMB's series of measures, the most valuable thing is not the efficiency figures but the re - definition of "talent." Shu Lan introduced that in the past, the bank mainly recruited people majoring in finance and accounting. Now, a certain proportion is set aside specifically for recruiting technology talents in the fields of information technology and computer science.

This transformation of the talent structure is not without resistance. In fact, it has created a subtle "cognitive gap" within the bank. On one hand, there are technology newcomers who master programming thinking and are good at communicating with machines. On the other hand, there are traditional financial people who have been deeply involved in the business for many years and rely on experience for judgment. Their situations are very different in the wave of AI.

When AI starts to replace basic copywriting and standard customer service, old employees without a technical background inevitably have career anxiety - worrying that the experience they have painstakingly accumulated will be completely deconstructed by algorithms overnight.

However, in Shu Lan's view, this kind of worry misses the point. She believes that what AI exposes is not "redundant manpower" but a "capability gap."

The impact of technology is essentially forcing bank employees to re - examine their core competitiveness. If a person's value only lies in executing standardized processes and delivering fixed scripts, being replaced is only a matter of time. But if he can use AI as a lever and use the saved time for in - depth thinking, complex negotiations, and risk insight, then AI will become his most powerful assistant.

That's why Shu Lan repeatedly emphasizes a point: The AI competition in the bank is, on the surface, a competition of technology investment, but at a deeper level, it is a competition of human quality in "who can use AI better." Based on this underlying logic, when the outside world focuses on the sensitive word "lay - off," her answer is particularly calm: "Whether the bank lays off employees or not depends on the economic environment, not AI."

In her view, AI is more like a powerful tool that can release human resources rather than simply eliminate positions. There are two types of diversion within the bank. On one hand, for those basic marketing and customer service positions that are repetitive and do not require much creativity, if employees' abilities cannot be improved, they will indeed face adjustment. On the other hand, employees with comprehensive qualities will be transferred to core positions that require more human judgment.

"AI can only extract data, but the banking industry deals with people and enterprises. In the process of communication, that kind of professional judgment based on experience, intuition, and even emotion is something that AI cannot participate in," Shu Lan said.

05

Conclusion

Four levels, one main line. Through these cases, a core law gradually emerges: the depth of integration between AI and business directly determines whether it creates chaos, simply improves efficiency, or reshapes capabilities.

From another perspective, the way enterprises treat AI ultimately reveals their attitude towards "people": Are people regarded as executable units that can be monitored and replaced, or as the core of value who can master tools and make complex judgments? Different answers lead to completely different results brought by AI.

True transformation begins with letting go of illusions. In an era when AI is becoming increasingly good at providing definite answers, the ability to ask good questions, make complex judgments, and master tools rather than being mastered by them is becoming as precious as gold - that may be the greatest truth in the AI era.

This article is from the WeChat official account "DoNews" (ID: ilovedonews), written by Yan Qiu, and published by 36Kr with authorization.