Author | Teng Xinyang
Curated | Luo Yanshan
As the hottest trend in the current technology industry, the influence of artificial intelligence in all walks of life is heating up rapidly. In the face of the onslaught of the AI trend, some people see a vast opportunity, while others realize the challenges it poses to traditional models and methods. In order to meet the changes brought about by AI, the industry undoubtedly needs the support of a higher-level talent team, which puts forward higher requirements for the training industry.
How can we respond to this demand, and how can practitioners in the training industry grasp the opportunities during the changing times? A few days ago, Teng Xinyang, Director of the Shanghai Branch of the Human Resources Department of Neusoft Group, was specially invited to deliver a keynote speech entitled "How to Seize New Opportunities in the AGI Era" of the DTDS Global Digital Talent Development Conference jointly organized by Geek Time Enterprise Edition and Training Magazine.
This article is compiled from his talk and has been edited by InfoQ with no change.
Neusoft Group is the first company in China to obtain PCMM (Human Capital Maturity) Level 5 certification, and the company's human resources system has a large number of system documents and materials, forming a very mature knowledge base. For AI, the accumulation of knowledge base and data is very important, so when the AI boom began last year, we immediately anchored this technology and began to think about how to make good use of AI's capabilities.
Just as playing cards can be used as teaching aids, magic props, or winning tools, tools are important, but whether the people who use them can make them valuable is more important. Neusoft has encountered a lot of ups and downs in the process of doing AI, and I have learned a lot of people and a lot of ideas, and I also summarize my feelings in three sentences:
1. Instead of predicting tomorrow with yesterday, focus more on thinking logic than on the presentation of the sharer's content.
2. When the boulder cracks, some see fear, and some see light. I hope that you can find your own value proposition and opportunities through my sharing.
3. Weakness and ignorance are not obstacles to survival, arrogance is. We need to look at the application scenarios that I will share next with a growth perspective.
The content of this sharing revolves around three keywords, that is, in the era of generative AI technology, we "Why", "What" and "How", and the core foothold is "Change". AI is different from some other concepts that have been hyped up in the past, and its popularity has continued since last year, indicating that it must have brought about some fundamental changes, which leads to our first topic.
Why is the era of generative AI changing?
Value is a chain, which is fluid, and is also divided into surface value and underlying value. What is surface value? For example, the person who studied how BB machines could be interconnected with mobile phones before mobile phones were popular, and later technological developments proved him wrong. So if you only think that AI can help me do classes, be a digital human instructor, I think this is the surface value, and we should pay more attention to the bottom value: why should I use AI to do classes? Why do businesses need a large number of courses? Pay attention to how the essence has changed.
Looking back on the history of human resource development, in the early stage of capitalist development in the 17th century, whoever has money can afford more family members to work for themselves, I only need to provide food and housing, and the business can be very large, but after entering the era of electrification, machines began to replace manual work, so I need more non-family members, and human resources began to emphasize scientific management and pay attention to the output efficiency of personnel. At this time, white-collar workers also appeared, and the concept of strategic human resources was born in order to manage their output.
In the era of informatization, scale has begun to be replaced by customized and economical services, and the market is highly segmented. That's when we start to pursue value and weed out those who might not be doing well. So in the future, with the advent of the era of artificial intelligence, how will we change?
First, a combination of human-robot collaboration may replace or augment our existing workforce and labor relationships, which is why HR must focus on AI technology. Second, we will have high-impact human resources organizations, and human resources will be pan-functional, and enterprises will need more roles such as AI designers of human resources. In fact, in each era, the change of human resources represents what was lacking in that era, and in the mechanical age, it pursues efficiency, in fact, it represents the lack of good processes; In the era of electrification, it pursues quality, and what it lacks is good leaders; In the information age, we pursue value, and on the other hand, what is lacking is data, because data can accurately quantify value. In the future era of AI, I personally think that the pursuit of innovation may mean that we lack some new quality productivity in this era.
Extrapolating from the relationship between cause and effect and time, you will find that we are forced, never active. So why should we focus on technology? Because there is a causal chain between technology and us. New technology is not the first to change human resources, but first to change the enterprise, the business model, with this business model, there will definitely be people to expand it, I hope that more people will come to help me make money. Therefore, after he has an organization, he has a requirement for talents, and human resource management will naturally be affected and begin to change. Therefore, there is a chain relationship between technology, organization and talent.
On the other hand, HR tends to be on the cusp of the wave in the process of technology iteration, so we need to do two things, the first is that your mind should be on the cusp of the wave, and the second is to learn to surf and land smoothly before the wave hits down. That's what we're talking about why change.
What will change?
What do we say next? We just mentioned the business model, organization, and talent of business organizations.
business model
So, first of all, will the business model change? McKinsey and Accenture here are two reports that tell us that the percentage of companies adopting AI technology in at least their business areas has increased from 20% to 50%, and there will be accelerated change in the future; At the same time, the share of AI-driven revenue has doubled from 12% to 25% in '21, so it is likely that the existing business model of enterprises will change dramatically due to the arrival of AI.
Organizational structure
Will there be any changes to the organizational structure? Historically, the most important change in the information age has been the migration of computing platforms. For example, IBM launched the first personal computer, the 5150, and its interaction mode was the drag and drop of the mouse. In that era, you will find that SAP and Workday have put forward the concept of EHR, called human information system, that is, the management of employee data by category.
For example, if my company's information system extracts an employee's file, can I know how old he is? Is it male or female? What major? Where do you work? What position do you work in? What is the monthly salary? But if I ask next, do you say he likes to play basketball or does he like to play football? What kind of leaders does he associate with? In what business scenarios is the turnover rate high? Maybe this system won't give you the answer. Therefore, the EHR system is divided into large categories of data management, creating a department-level organization, and many organizations in that era were department-level, and the information barriers between different levels were very strong.
The birth of the iPhone marked the evolution of the computing platform from the computer to the mobile phone, and the change it brought was the refinement of data. The mobile phone can know your GPS location, can know your fuzzy search and then recommend you to achieve personalization and other needs. So in 10 years or so, the concept of DHR came along, which is to be based on more breakdown data that allows you to know your people better and manage them. In other words, you should know your employee's personality, what kind of leadership he likes, in what kind of environment he has high performance output, and even what skills he lacks.
Personally, I think it will create the concept of a diamond-shaped organization. In the previous section-level organization, there were countless subdivisions under the middle and senior levels; The diamond-shaped organization is more focused on the vision above, someone tells you that we can make money in this direction, and there will be RPA to replace your repetitive work below, and then use AI to provide some recommendations and suggestions for your selection, rewards and punishments. For example, if a company encounters a problem on the road to finding money, I think I can do it, then I will be PM, and then I will tell the AI, you recommend me in the enterprise, I need 3 programmers, two finance, legal affairs, and recruitment, please help me recommend suitable and qualified candidates to me, form a team, and after solving this problem, the team will be disbanded, and the personnel will wait for the next project. This is called a rhombic structure.
In the era of the second-generation computing platform, it has been accepted by many native Internet manufacturers. The characteristic of the diamond-shaped organization is volume, and if you are not invited in the organization, either you can't live, or you have bad popularity, you will be brushed down by this organization, so it naturally has the characteristics of volume.
The third computing platform, I think the hallmark is the release of Vision Pro last year, the metaverse, virtual reality is very likely to bring about a DAO (decentralized organization) change to our organization. What we can see is that the individual proportion of employees is getting stronger and stronger, and the framework management of the organization is getting weaker and weaker. From the section level to the diamond shape, you will find that super employees are more likely to stand out in the organization.
Productivity structure
Will there be a change in the structure of productivity? OpenAI previously released an industry exposure report, and everyone joked that you will find that the exposure of manual workers is very low, and the exposure of white-collar workers is particularly high, such as writers, interpreters, etc., then those of us who sit in the office will be replaced by AI. But in fact, we believe that the concept of exposure is not substitution, but enhancement, that is, the roles and positions with high exposure are likely to have the concept of super employees in the future, and the probability of super employees will be particularly high. This means that the company cultivates talents, which may be a disguised cultivation of super competition, and in the end, he goes out and adds AI, and three people set up a company to grab orders and customers with you.
In the future AI era, 20% of humans and machines may replace 80% of repeatable, routine, and structured work, while the remaining 80% of people will rely on human-machine collaboration to complete 20% of human-related, unstructured work, and this part of the work will consume a lot of time and energy. So where will it be safer in the future when AI arrives? We're looking for enhancements, not replacements.
human resources
Will the mission of HR change? We mentioned a few key words, the first two are "efficiency growth-oriented" and "efficient working environment", which are the core of traditional human resources, without going into details. The last three keywords are differentiation and differentiation.
First, it's important to be able to create a digitally inclusive workplace. The biggest obstacle here is the middle level, because the boss can now know what the employees are like, and they don't need to know about it through the managerial level, so the middle level will disappear, so it is the biggest obstacle. The key point is also on them, and whether they can create a digitally inclusive workplace is important.
The second keyword is personalized employee experience creation, and we just mentioned that digital HR is talking about data segmentation, and experience creation is about helping you mine better, cleaner data.
The third key word is to focus on the sustainable development of employees. With the advent of the era of super employees, you don't want to teach him for two years, and he will go out to be your competitor, right? So in this case, we should pay attention to the win-win situation of the enterprise and employees.
Last year, Deloitte released a report called Future Human Resources, which mentioned that the core of management should be from the organization to the people, sustainable development and agility, global learning and process optimization, the pan-functional structure of our HR, and the experience architecture to release value and productivity. Especially now, when we are facing Gen Z, if your experience architecture is not done well, it is very likely that employees will earn more than you work part-time after work, so why should I put energy and value output in your 8 hours? Finally, there is the use of human-robot collaboration to improve quality and speed.
Now let's look at the term "new quality productivity". There are three elements of productivity: laborers, means of labor, and objects of labor, which will change in three directions: higher quality, higher technology, and wider scope.
Similarly, for human resources, the reform should also start from the three points of labor, labor relations, and labor materials. Taking Jingdong warehousing as an example, the efficiency of warehousing in the past depended on how fast the little brother ran, because at that time, it was all pushing carts, and how fast he ran determined the efficiency of your order. At that time, there was also a big screen, that is, someone looked at the screen and took a microphone and shouted, what number, you are running slowly, run faster.
Later, these warehousing brothers began to learn Java and development one day, and then it became our today's black light factory, which is JD.com's new quality productivity. Its laborers have changed from people to machines, and the means of labor have changed from efficiency data to profit data. In the past, I was pursuing efficiency, but now I don't look at efficiency from the big screen, I directly look at how much the turnover has increased. The object of labor has changed from goods to customers, and he has changed from focusing on the efficiency of goods to focusing on the experience of customers, so his entire new quality productivity has been greatly improved. Then we should also pursue new qualitative productivity through these three frameworks.
How to change?
Key talent
How will this last part be discussed? We set up the Apollo Innovation Lab in April last year, which is made up of a few very interested HR friends, all of whom have pure HR backgrounds, which means they don't know how to code. So you don't have to worry about whether people who can code can continue to do AI research and development in HR, which is not the case. We are doing one thing, called the Digital Life R&D Project, which is to create some digital employees.
Combined with the framework we talked about earlier, let's first look at workers, that is, enterprises have defined two types of talents in the transformation process of HR AI, the first category is called important talent, and the second category is called key talent. Key talents are the core talents to build differentiated competitiveness. For example, the pilot of an airline is an important talent, and the plane cannot fly without him. But for example, Spring Airlines focuses on cheapness, and its differentiated competitiveness comes from operational data management, while China Eastern Airlines focuses on service, and its differentiated competitiveness comes from flight attendants, so this talent is the key talent.
Therefore, we internally divide the talent team of HR AI transformation into two categories: key talents can complete AI independently, and are responsible for the innovation and creation of application scenarios, and may not need to have too strong experience in business scenarios; The key talents are those who are very experienced, who may not understand AI, but who can quickly grasp the pain points in this business scenario.
We have also come up with an innovative concept called atomic-level substitution. When you disassemble the existing human resources activities in the enterprise into atomic-level activities that can no longer be disassembled, and then see if this atomic-level activity can be replaced or enhanced by AI, this is actually the essence of innovation. We looked at what generative AI can do, what kind of work is essentially replaced, and in which atomic-level activities will that work be replaced or enhanced? (Actually, the concept of substitution and enhancement is very simple, 100% is called substitution, 80% is called enhancement.) And then these atomic-level activities are in turn requisitioned by which upper human resource management activities?
Internally, we've now tapped out more than 200 atomic-level activities, some of which are substitutions and some of which are enhancements. We're still working on HR's own exposure formula, so if there are 10 atomic-level activities and 8 of them will be replaced by AI, then you're 80% enhanced.
Key Scenarios
After emphasizing key talents, let's talk about key scenarios. When we are looking for key scenarios, we must have critical and innovative thinking about existing applications. It's not that this thing is valuable when you take it and use it, or that most people have value when they use it this way. We've also identified some key scenarios and enumerated HR's digital workforce. The first part of this system is called Jarvis, which is an AI-built personal super-assistant that collects more data about employees and allows us to understand them better. The second one is called Bodhi, which is a learning assistant for employees. The last one is called Magic Mirror, which is an AI recommendation assistant that can recommend who is suitable for what position and who is most suitable for promotion, and provide us with comprehensive advice and guidance.
This time, we will focus on Bodhi, which is a learning assistant, that is, the enhancement of knowledge, and the active push of search results, not to let people find knowledge, but to let knowledge find people. It comes from our substitution and critique of existing learning tools or learning methods in the enterprise.
Under what circumstances do employees like to learn? In the first case, I encountered a problem, and I knew that learning skills could be improved in a short period of time, which could help me solve this problem. But what is the solution for you? He wants to say that you don't know how to solve something when you encounter something, and when your eyes are dark, I hope you can accurately search for the content you want to learn in the search box of the company's online learning platform, do you think employees can do it? Second, when I first joined or was promoted, I wanted to be able to reach a higher level through my studies. Then the solution given by the company is to throw you the route or job position system for vocational skill promotion, and as a result, for employees, as the company grows older, this system will become more and more complex. From the perspective of employees, forget it, I'm not promoted, it's too complicated. So you'll find it to be a state of high cost investment and low return on learning.
So what do we want Bodhi to do? If you come across something, I want you to tell me about it, and I'll tell you how to solve it, and then at the same time give you advice and tell you what kind of course to take. After completing this course, I will evaluate you to see how well you have mastered the skills. If you have this certification, you may be retrieved by the magic mirror in the future, and you will be able to earn income from solving the corresponding problems.
Second, if you want to be promoted, we don't want to tell you how to be promoted based on the ability of the position, but according to the concept of skill system and skill tree, if you want to make more money, you have to produce greater value, so what kind of skilled employees can output such value? This point is inspired by the hints of the game, there will be skill talent trees in the game, Xiaobai points different talent points may become a warrior, mage, he knows his route very clearly, and can even plan his route in advance. I certainly want our business to do the same, so we want to build a skill tree.
For example, Java is a skill tree, if Java is divided into 100 knowledge points, then each knowledge point is classified from level 1 to level 10, and then 30 questions are produced at each level, and employees may only need to answer 10 questions, and I can know what level your skill point is, for example, you have reached level 4. Next you want to get promoted, if my requirement for you is level 10, give you half a month, you have to be promoted to level 10. Next, Bodhi will push you a course, and you don't have to start reading this course from scratch, you can start reading it from 40%, because Bodhi already knows that you are at level 4.
You will find that in the era of AI, this thing is very simple, in the past, it took a lot of experts to study for a long time and spend a lot of money to come up with a problem and grade, but now a large model with world knowledge can quickly give a solution. As a result, the cost of this matter has become very low, and the decision-making based on training has become clearer and simpler. In the past, there may have been bosses who pat their heads to make decisions, such as those who want to be All in AI, because they are forward-thinking, but they don't have a clear understanding of the current situation. If your skill tree tells him that the talent pool for our AI skills is red, not green, he won't make that decision. He will definitely mark the skill points around the red skill first, send these people over to train the AI, and then turn the light green after learning it, and then make the strategic decision of All in AI. So it's actually a better starting point for managers.
We conducted experiments on Bodhi and found that the difference between the accuracy of AI recommendation and the accuracy of human recommendation was less than 10% in the scenario of class promotion.
In other words, what AI pushes is no different from what people push. However, this difference decreases as the number of courses increases, and the same applies to our Magic Mirror test. 2,000 courses, with the same accuracy as human and AI recommendations; With 20,000 courses, AI pushes more accurately than humans. This leads to another topic, what should I do if I don't have enough courses in my company?
We are now training Bodhi, and in three days we searched 170,000 courses on the whole network, which may all come from some open source communities. Then we now let Bodhi himself reduce these courses to 3,000 courses, and then let human experts score them to find out which ones Neusoft needs.
So you will find that the trainer has a very important point in this process, and with Bodhi, your training manager may know what kind of technical obstacles your team is encountering before the project manager of the front-line business. For example, if all members of your team are searching Python and blockchain technology per unit of time, it proves that your customers are having such problems. Therefore, whoever grasps the tipping point of the data and analyzes it will have a greater say.
Training enhancements
Finally, let's talk about labor materials. We've been investing in interactive gamification for a long time. You'll find that data is nothing more than three steps, data mining, labeling, and analysis, and now AI has an additional training step.
The best thing to do in this process is the game, the game is the best to understand these players, if the glory of the king wants, it can know your investment risk appetite, you can know what kind of girlfriend you like, because you have a lot of behavioral data recorded in this game. That's why we've done a lot of gamification innovation and developed our own e-sports knowledge training platform. We experimented not only at home, but also in many Fortune 500 companies. According to the data we obtained, in the new employee training, the active knowledge answering rate of employees of a Fortune 500 company on the e-sports platform has been greatly improved.
With that amount of interaction, I can know exactly what your skill points are, and then we do some behavioral analysis of these people to understand their behavioral preferences and appropriate development directions. For example, if a programmer has a lot of social behavior in the game, then he will be suitable for pre-sales consulting in the future. A lot of social-friendly employees come from certain universities, so if your business needs such employees, you can go directly to these universities in the future and recruit people. And all of this comes from a five-day training session for new employees. You'll find that the impact of training is greatly enhanced, and you can influence HRBP's recommendations in the future, and you can influence the recommendations of hiring decisions in the future.
brief summary
That's all I have to say. There's a saying I love, "Change is slow, but it's done in a split second." I also expect that when we increase our number of digital employees to 15-30 this year, we will be able to directly reduce the cost by 1 million. This is actually quite a big challenge, and I hope to be able to share more with you next time, thank you.
Original link: Neusoft Group: In the era of generative AI, how to lay out AI human resources strategy? _Featured articles by Generative AI_ Teng Xinyang _InfoQ