AI-Powered Management: Embracing the Generative Approach to Corporate Strategy

  In November 2022, the American company OpenAI launched the Generative Pre-training Converter (ChatGPT) for human-computer interaction, making the concept of artificial intelligence generative content (AIGC) sweeping the world. Generative artificial intelligence is becoming more and more popular, not only affecting the way individuals use information, but also gradually reshaping all aspects of corporate management.
  Taking corporate strategic management as an example, the traditional corporate strategic decision-making process is based on points and areas, centered around the cognition of the company’s top managers, forming a top-down empirical management model. However, as the business environment becomes increasingly complex and industrial changes gradually deepen, the development direction of enterprises takes the form of multi-directional radiation. Experience-based management based on the individual cognition of senior managers is often difficult to effectively deal with complex internal and external factors, and may even cause the enterprise to fail. Trapped in the manager’s cognitive comfort zone and unable to respond to the turbulent external environment in a timely manner.
  Today, when artificial intelligence applications are gradually becoming more popular, some manufacturers have begun to create management assistance tools for specific industries or fields based on artificial intelligence technology. They are trying to use artificial intelligence to help business managers break through cognitive boundaries and build a generation system that integrates human experience and artificial intelligence. management decision-making model.
Generative artificial intelligence: from uncertainty to certainty

  Generative artificial intelligence is an artificial intelligence neural network that is based on huge amounts of data and is formed through preset and self-correcting feedback rules. It can generate higher-level data. It can understand the elements of human creation and recombine these elements to create new content that can be understood by humans. An article published by the OpenAI team in 2022 explains in detail the three shaping stages of generative artificial intelligence: strategy construction, scoring ranking and decision optimization.
  In the strategy building phase, the research team used data labeled by humans to build a pre-trained neural network model. In the scoring and sorting stage, artificial intelligence formulates different result scoring strategies on its own, and assigns and sorts multiple output results generated based on the same task instructions. Human trainers use personal experience to guide the artificial intelligence in choosing the best scoring mechanism. In the decision-making optimization stage, the artificial intelligence neural network generates different content at high frequency based on the output strategy, and uses an internal scoring mechanism to assign scores to massive content, optimize the output strategy on its own, and finally give the best decision to meet external instructions.
  As can be seen from the above description, the basic working principle of generative artificial intelligence is to first generate multiple results that may match the instructions based on the output strategy, and then gradually increase the possibility of generating strong matching results based on the scoring mechanism until the matching degree is the highest The only solution was born.
  From generating possible results to judging and reducing the possibility of results, generative artificial intelligence is gradually transforming the output content from uncertainty to certainty. Figure 1 shows the entire process from complete uncertainty to certainty.
Uncertainty, discontinuity and types of business management

  Currently, generative artificial intelligence is mainly used in language-related fields, and is therefore often called a large language model. In the big model of generative language, you do what you say. As long as the computing power is sufficient, the process of this model finding possible results in massive information, judging the possibility of the results, and obtaining deterministic output based on this is continuous and guaranteed.

  However, the reason why the popular application of artificial intelligence has just begun is because the huge algorithm, computing power and data gap between the theory and practice of artificial intelligence has only recently been broken through. The stock price of NVIDIA, which played a key role in this breakthrough, has been soaring in recent times, and it is still hard to find a card globally. Many startups and even major Internet companies have expressed their desire for All-in generative artificial intelligence, but few can actually achieve it. This also reflects the difference between knowing and doing, and the discontinuity between the formulation and execution of corporate plans.
  If we sort out the relationship between “knowing” and “doing” from the two dimensions of uncertainty and discontinuity, we can get Figure 2. The vertical axis of this figure represents uncertainty in “knowing” and the horizontal axis represents discontinuity in “doing”. “Knowing” is the recognition of the future development direction, and the main challenge is uncertainty. “Action” is the practice on the future development path, and the main challenge is discontinuity. Uncertainty about future development directions can be divided into high and low situations. At the same time, whether we can achieve the unity of knowledge and action from now to the future depends not only on the certainty of “knowledge”, but also on the continuity of “action”.
  There are many paths from now to the future. Which path to choose is determined by “knowledge”. The improvement of “knowledge” from uncertainty to certainty can help people choose the path they think is the best, but whether the path in reality is To succeed, we need to overcome the discontinuities on the future path.
  Based on Figure 2, we can divide the enterprise management model into four types along the timeline of enterprise development: adaptive management, visionary management, planned management and emergent management.
  When the uncertainty of future development direction is high and the discontinuity of future development path is high, the development of things is in a state of chaos. Most companies adopt adaptive management and eliminate many development directions in an attempt to find a suitable resource and capability for themselves. The development direction of conditions, and bet limited resources and capabilities on this development direction, striving to achieve a breakthrough in the development path and enter the growth stage.
  Companies in the growth stage often adopt vision-based management. There are always companies that can overcome the discontinuity in the development path, realize their own vision, enter a stage with low uncertainty in the development direction and low discontinuity in the development path, and begin to adopt a planned approach. Management model, try your best to stabilize your development direction and continue your development path. However, when companies face drastic changes, they have to make new choices to adapt to the external environment with increased uncertainty in the direction of development.
  From this we can see that at different stages of the life cycle, enterprises need to adopt management models that are suitable for uncertainty in development direction and discontinuity in development paths. If we divide the enterprise life cycle into four stages: entrepreneurship, growth, maturity and transformation, then the enterprise will adopt four management modes: adaptive management, visionary management, planned management and emergent management at the corresponding stages.
Generative artificial intelligence management thinking

  Today, as the business environment becomes increasingly complex and the pace of enterprise development continues to accelerate, an enterprise may go through the four stages of entrepreneurship, growth, maturity, and transformation in just a few years, and enter a development cycle of secondary entrepreneurship, growth, and maturity. middle. These companies will face four combinations of uncertainty in development direction and discontinuity in development paths at the same time. The various businesses of the company will also be divided into four types: innovation, growth, maturity and transformation. In such a complex situation, how companies should coordinate various businesses and how to grasp the emerging development directions and changing development paths in a complex environment are major challenges faced by corporate management decision-making.

  In the Internet era where information is diversified and easily accessible, planned management can easily cause companies to ignore abundant external signals, leading to organizational rigidity and missed development opportunities. Too many market signals will also cause companies that adopt emergent management to wander randomly and spread disorderly in the market, seeing multi-directional development paths but being unable to follow them consistently. Therefore, companies that lack the ability to process massive amounts of information at high speeds will find it difficult to create optimal performance regardless of whether they adopt adaptive, visionary, planned, or emergent management. Facing the complex external environment and diverse business portfolio, enterprises can learn from the logic of generative artificial intelligence technology and try generative artificial intelligence management logic.
  The basic logic of generative artificial intelligence management is from divergence to convergence, that is, from “birth” to “achievement”. In the divergent stage, the generative artificial intelligence management system obtains massive information from various businesses and pre-trains the enterprise decision-making model, so that it can learn the laws and patterns in the data, and then predict the enterprise’s subsequent business direction and future conditions. In the convergence stage, the trained model enters the decision-making assistance simulation stage, and the data generated are for managers’ reference. Managers also evaluate the decision-making options output by the model in order to subsequently improve the model and improve the quality of the generated results.
  In practice, the decision-making approach adopted by enterprises that combines bottom-up and top-down has the shadow of generative artificial intelligence management. For example, when discussing the transformation of Huawei’s decision-making system, Ren Zhengfei said: “I don’t know how many resources are needed on the front line. I can only let the person who can hear the gunfire call for gunfire, because he is closest to the customer, everyone listens to him first and chooses first.” Believe him. When we review the review afterwards, we find that we have wasted ammunition, and then we will ‘settlement the accounts later’ and sum up the experience.” Every call for fire from Huawei’s front line is a bottom-up strategic output, and the subsequent review and reckoning by the management are also one Top-down revised returns. The goal of artificial intelligence is to label a large number of correction returns and establish an institutional foundation for the organization to achieve self-training of “output → feedback → adjustment → output”.
  When an enterprise develops along its own life cycle, from the entrepreneurial stage to the growth stage, and then enters the maturity stage and transformation stage, the uncertainty of the future development direction faced by the enterprise at each stage is combined with the discontinuity of the future development path. Although each is different, the combinations at each stage are similar. There are also rules to follow in the changes from one stage to another. In this case, enterprises can make management decisions according to the thinking modes of adaptive management, visionary management, planned management and emergent management.
  However, with the continuous development of enterprises and the rapid changes in the external environment, many enterprises have a combination of innovative business, growth business, mature business and transformation business at the same time. These businesses face different uncertainties in development directions and discontinuities in development paths. In order to cope with the opportunities and challenges faced by various businesses, enterprises will serve different types of users, adopt corresponding organizational structures, develop products that meet user needs, and try to compete and cooperate in different market environments. As a result, the complexity of business management began to increase exponentially.
  From the corporate headquarters level, each business faces different specific situations and cannot adopt unified management mechanisms and methods for management, making it difficult to achieve synergy between businesses. From different business perspectives, due to the different specific situations faced by each business, it is often difficult to meet the requirements of the company headquarters, and it is also difficult for each business to form synergy with other businesses.

  This situation can be explained by Figure 3. The base is a differentiated combination of uncertainty in the development direction and discontinuity of the development path faced by the business at different stages of development. Vertically, as the complexity of the business increases, the difficulty of integrating knowledge and action increases. In this case, enterprises can consider drawing on the working logic of generative artificial intelligence technology and applying the generative artificial intelligence management model.
  The generative expression includes two keywords, one is “生” and the other is “成”. The basis of the overall management objectives of the enterprise is the management objectives of each business, and the management objectives of each business include user management, organizational management, product management and market management. Due to differences in the development stages and specific development situations of different businesses, these businesses will generate massive amounts of information in actual operations. Enterprises can use pre-trained management models to label the management problems and data information faced by each type of business, and continuously optimize the models through the evaluation of managers to form a bottom-up management information collection system.
  The second key word of generative artificial intelligence management is “success”, including the achievement of various business goals and the achievement of the company’s overall goals. Because the management goals of each business include specific goals such as user management, organizational management, product management, and market management. Therefore, the collection of these goals should achieve the overall management goals of each business, and the collection of the overall goals of each business at different stages of development should be able to achieve the overall goals of the company level. Corporate senior managers can evaluate and rank the achievement of each business goal from top to bottom, and adjust the company-level return model, thereby helping each business achieve its own goals while helping the company achieve its overall management goals.
  In the generative artificial intelligence management model, in addition to the two keywords “生” and “成”, which represent bottom-up and top-down management processes respectively, “artificial” and “intelligence” are also important Key words. “Intelligence” emphasizes that business management is a science that can deal with uncertainty in management through the formulation of rules. Use the certainty of rules to deal with the uncertainty of change, and use the continuity of processes to deal with the discontinuity of behavior. The keyword “artificial” emphasizes that management is an art or even a craft. Managers are required to conduct moderate intervention in the management system, including but not limited to evaluation of management data, feedback on management actions and optimization of management models.
Case studies from Nut Power

  Art pursues openness of results and science requires clarity of results. Management decisions using generative artificial intelligence need to combine openness and clarity to achieve a balance between “generation” and “becoming”. The following takes the gaming industry as an example, selects the development case of the well-known Chinese game manufacturer Nut Power, and traces the historical process of incorporating the American artificial intelligence company into the core of decision-making, summarizing the decision-making context of generative artificial intelligence management: wide-area exploration , distributed mining, vectorized derivation.
  First, simplify wide-area exploration based on competing product tags. The gaming industry is known for its rich content and high level of competition. According to official statistics from Valve Group, the “Steam” electronic game distribution platform developed by it will release a total of 12,857 new games in 2022 alone. On the one hand, the abundant market competing products provide game manufacturers with diverse market experience for reference and imitation; on the other hand, the dazzling new world also causes product developers to get lost in the exploration of the unknown, causing product development to lose focus. Losing the core of the product causes players to lose their desire to buy.

  Therefore, game manufacturers face a high degree of uncertainty when developing new products. They neither know what they can do (possible results) nor what they are best suited to do (possibility of results). Based on the evolution model of “from uncertainty to certainty”, when developing new products, game manufacturers need to first choose a track that suits them.

  Generative artificial intelligence forms a strategy model based on labeled data, and outputs input information into multiple possible results. Game manufacturers can also refer to this method for product development selection, using hierarchical tags to summarize a large number of market competing products, simplifying the cost and difficulty of industry-wide searches. Qiang Shuzeng, the founder of Nut Power, said that as early as 2015, he recognized the role of in assisting its product development, and actively tried to connect with the founding team of As an artificial intelligence company focusing on the mobile market, especially the mobile gaming market, has been trying to digitally classify products on the market since its birth.’s flagship function App IQ provides a system application taxonomy based on the mobile application industry. Its tagged content is not limited to the simple classification of mobile applications by software publishers such as App Store and Google Play, but also deeply mines application content and provides application-level category tags in 19 categories and 152 subcategories.
  The founder’s keen awareness and advance layout of the prospects of artificial intelligence applications, as well as the category-feature dual-level label classification provided by the artificial intelligence assistance system, enable Nut Power to quickly, concisely and not overly generalize scan the entire gaming industry products. become possible. This allows Nut Power to quickly position its products and identify the tracks on which it wants to develop.
  Second, match R&D capabilities and promote distributed mining. After roughly clarifying several tracks that may match the company’s capabilities, Nut Power relies on artificial intelligence to evaluate the future development space of each potential track and conducts rankings to further reduce the uncertainty of market information. Generative artificial intelligence achieves sorting of outcome possibilities by optimizing the scoring mechanism, and the key to this step is to output multiple solutions based on the strategy model. In the context of game product development, it means multi-product trial and error iteration and direction optimization selection.
  Relying on early manual data input and subsequent intelligent data monitoring, Nut Power effectively uses artificial intelligence to guide the research and development direction of the internal team. Nut Power has rich experience in product trial and error iteration. The company has launched a total of 146 products in the Apple and Google stores, of which there are 62 active applications as of June 2023 and 84 applications that have been removed from the shelves. These active applications have sustained and stable daily active users and daily revenue.
  Third, through content migration, vectorized derivation is achieved. The key to the maturity of generative artificial intelligence is the mutual intervention of decision-making models and scoring strategies. It enters a self-optimization process without human trainers. When receiving external instructions, it can self-derive massive content and quickly sort it to select a unique solution. In the same way, the purpose of real-life enterprises to comply with market supply and demand relationships is also to launch products that meet consumer needs, gradually understand market needs in the process of being supervised by the market, and ultimately achieve proactive prediction of market trends. Nut Power uses’s artificial intelligence data to focus the company’s attention on the integrated category of “operation + adventure”.
  The company refers to the content-level feature tags of, compares the content designs of competitors in the same category, and selects the elements and features most relevant to scene adventures and farm operations, such as character development systems, puzzle treasure hunts, and plot decisions. Subsequently, through the mixed coordination of new and old employees in the planning, art and programming groups, the multi-category game development experience accumulated in the past was applied to the focused product development, and the product was transformed from the 2019 hit product “Polynesian Adventure” to the latest hit A spin-off of the game “Sarah’s Adventure: Time Travel”.
  The development of generative artificial intelligence technology brings not only challenges but also opportunities. Companies that actively embrace generative artificial intelligence technology can strengthen existing and even gain new competitive advantages. Although generative artificial intelligence has not been around for a long time as a general technology, its problem-solving logic has been applied by many companies. Entrepreneurs who continue to actively apply generative artificial intelligence thinking to reshape management will be better able to Find certainty in certainty and create continuity in discontinuity.

error: Content is protected !!