Artificial Intelligence and Management: Evolution, Applications, and Limitations

(1) AI Will Disrupt the Way Core Management Activities Are Conducted

AI, especially AIGC, not only brings improvements in management efficiency but will also enhance management quality by changing the way managers work and how organizations operate. The impact of AI on management activities includes the following aspects.

1. Decision-Making

AI can quickly make accurate predictions and decisions by analyzing vast amounts of data, providing insights and understandings that exceed human capabilities, thereby helping managers make better decisions. For example, AI can generate precise financial forecasts based on historical data. In addition, AI can create complex financial models to help financial decision-makers simulate the potential financial impacts of different decisions.

2. Task Automation

AI can automate more basic, repetitive management tasks. For instance, enterprises can optimize financial processes by automating billing, invoice processing, and other financial tasks, reducing human errors and improving work efficiency. Particularly, AIGC can perform some “creative” tasks that previous AIs could not accomplish, automatically generating and creating reports and data analyses, allowing managers to obtain relevant information faster and reduce the workload of manual operations, thus freeing up their time to focus on more complex issues requiring human expertise and emotional intelligence.

3. Employee Management

AI can improve the recruitment, evaluation, and management processes of employees. For example, AI can be used to parse job applicants’ resumes, predict their job performance, and better monitor employee performance while providing instant feedback.

4. Performance Management

According to the latest research article from the MIT Sloan Management Review and Boston Consulting Group in May 2023, titled “AI Is Helping Companies Redefine, Not Just Improve, Performance” (Schrage et al., 2023), AI is helping companies redefine their performance, not just improve it. Effective governance of key performance indicators (KPIs) enables leaders to transform KPIs into sources of competitive advantage. AI algorithms can analyze the relationships between multiple KPIs and their underlying components to better balance competitive and/or complementary interdependencies.

5. Customer Relationship Management

AI application tools can help provide more personalized customer experiences, effectively targeting and predicting customer needs, enhancing customer satisfaction through better customer segmentation and pricing strategies. For example, AIGC chatbots can generate personalized recommendations based on user behavior and preferences, providing humanized customer service around the clock (Siegel, 2023).

6. Risk Management

AI can be used to detect fraudulent activities, predict market trends, and help managers better identify and manage risks. For instance, in financial decision-making, AIGC can assist financial decision-makers in assessing and managing the future financial risks of projects by analyzing past tasks and identifying patterns that may lead to defaults or credit risks.

7. Innovation Management

AI can identify and predict new market trends to drive the development of new products and services. AIGC can provide new thinking and perspectives, helping managers consider new possibilities in decision-making. This may contribute to the development of innovation and creative thinking.

Moreover, in the future, the number of AI digital employees will significantly increase. For instance, recent research by several scholars from Europe and the United States (Eulerich et al., 2023) shows that OpenAI’s latest generation AI language model, ChatGPT-4, has the capability to complete and pass the four major accounting professional qualification exams (Certified Public Accountant, CPA; Certified Management Accountant, CMA; Certified Internal Auditor, CIA; and Enrolled Agent, EA) with high scores. Undoubtedly, in the coming years, AI digital employees will have great potential in helping accountants address the widespread issue of labor shortages.

(2) AIGC Will Help Managers Make Better Decisions and Lead Organizations

Taking Microsoft Copilot as an example, we will discuss how AIGC will redefine human “work” and further assist managers in decision-making and leading organizations.

Microsoft Copilot is a new feature of Microsoft 365 that helps users complete various tasks more efficiently through AI technology, including writing emails, preparing PowerPoint presentations, searching for information, and organizing meetings. It can understand users’ intentions and provide relevant suggestions and actions. For example, in Outlook, Copilot can offer suggestions for writing emails to help users express their points effectively; in PowerPoint, it can integrate OpenAI’s image generator, DALL-E, to create custom images in text form; in OneNote, it can help users organize and format information; in Microsoft Loop, it can quickly summarize page content to help team members stay in sync (Microsoft Official Website, 2023).

The goal of Copilot is to create a new way of working. Satya Nadella, the chairman and CEO of Microsoft, recently expressed the hope that through the evolution of interaction with computing, work methods will fundamentally change, while releasing a new round of productivity growth. For management, Microsoft Copilot may bring the following transformations, as shown in Figure 1.

Here is an example of how Copilot redefines the financial management work of enterprises.

Suppose you are the financial manager of a company and need to generate a monthly financial report, including key indicators such as revenue, expenses, and profits. In the past, you might need to collect data from various sources and then manually input it into Excel or other financial software. With Copilot, this process can become simpler and more efficient.

You can directly send a message to Copilot in Excel within Microsoft 365, such as, “Generate a financial report that includes all sales revenue, operating expenses, and net profit from last month.” Copilot will understand your request, use your company’s financial data, and automatically generate a financial report in Excel. This way, you can complete all the work in one interface without switching between different applications, thus improving productivity.

Additionally, Copilot can assist you in analyzing financial data and providing insights about the company’s financial condition. For example, you can tell Copilot, “Analyze last month’s sales data to identify the products with the highest sales.” Copilot will understand your request, analyze your sales data, and then tell you which products had the highest sales.

These examples illustrate that Copilot can help financial managers complete their work more effectively, increasing their productivity, allowing them to devote more time and energy to financial analysis and decision-making.

Let’s take the work of an organizational manager, such as a Chief Financial Officer (CFO), as an example to explore how AIGC can help senior management make better decisions. In May 2023, the MIT Sloan Management Review published an article by Kristof Stouthuysen titled “How Digitally Mature Is Your Finance Office?” which lists seven key capabilities that CFOs should possess. We believe AIGC provides CFOs with new tools and perspectives, enabling them to better execute strategic decisions and leadership roles. CFOs should actively embrace and learn new technologies, understand how AI is changing financial management, and explore how to integrate these technologies into existing systems, leveraging AIGC to comprehensively enhance the following seven key capabilities.

1. Utilize Advanced Analytical Capabilities to Meet Strategic Needs

AIGC can provide more accurate market trend forecasts through deep learning and predictive analytics, helping financial leaders formulate more effective strategies. Additionally, they can simulate different strategic options and predict their potential outcomes, assisting CFOs in making more informed decisions.

2. Interpret and Explain Machine Learning Outputs

Machine learning models are often viewed as black boxes, as we do not fully understand how they arrive at predictions. However, it is now possible to attempt to generate better explanatory tools through more complex models produced by AIGC, such as automated feature importance analysis, helping CFOs understand how the models work and the implications of their results.

3. Data Management

The scale and complexity of data are growing exponentially, posing significant challenges for financial teams. AIGC can handle larger and more complex datasets, providing more powerful data management tools. For instance, automated data cleaning and preprocessing tools can help CFOs manage and utilize data more effectively.

4. Cultivate Analytical Skills

CFOs typically build data analysis capabilities within their financial teams rather than relying on experts. AIGC can help team members learn new skills faster while also changing the required skill set, such as using new AI tools and understanding more complex models. This may require more training and learning to ensure team members can effectively use these new tools.

5. Explore and Experiment

CFOs rarely succeed in utilizing data for analysis directly; they often need time to explore. AIGC may provide more experimental tools and methods, such as automated A/B testing and simulation tools, enabling CFOs to more effectively explore and experiment with new strategies and methods to generate high-value outcomes.

6. Manage Cultural Change

The conservative and cautious culture typical of financial teams may lead to resistance against new tools and methods. CFOs can change the communication and work methods of their teams by introducing relatively simple models of AIGC, ensuring all team members can effectively use these new AI tools.

7. Actively Participate in Digitization

CFOs need to ensure that everyone in their teams has the opportunity to access relevant training, technologies, and data, and learns new skills to unlock new opportunities. The emergence of AIGC necessitates further promotion and reinforcement of a data culture within organizations, including more data education and training to ensure all employees understand the importance of data and use it proactively and effectively to drive decision-making.

The future application and evolution of AI will occur simultaneously, with AI technology rapidly iterating, and its impact will be felt on both the demand and supply sides, inevitably causing comprehensive and continuous disruptive impacts across various industries. In the face of the AI wave, corporate executives should embrace AI with a positive attitude while maintaining a clear mind.

First, recognize that technology itself is not the source of competitive advantage.Reorganizing and simplifying business processes, carefully studying how large models can enhance operational efficiency and customer experience at various stages of end-to-end operations, further driving organizational optimization, operations, and management model reshaping.

Second, the innovative opportunities that generative artificial intelligence technology may bring.New technologies allow us to accomplish tasks that were previously unattainable, opening new spaces for survival and growth for enterprises, enhancing overall competitiveness in unique ways.

Third, applications in vertical fields.Enterprises can focus on developing their own vertical application algorithms using large models, extending and innovating based on their core business. An example of building a specialized large model in a vertical field is the application of AI in enterprise finance and accounting management (see Figure 2).

Fourth, human-machine collaboration.To fully realize the value of new technologies, at least at this stage, a collaborative work model between humans and machines is indispensable.

Fifth, integration with existing technologies.The adoption of new technologies and their integration with existing technologies will pose a challenge for enterprises, with the key being to establish trust between humans and algorithms.

Sixth, limitations and risks.Similar to many technologies in the past, the compliance, risk control, ethics, and regulation of AI technology lag behind its development, presenting certain uncertainties and challenges for enterprises in developing and applying new technologies (Karna, 2018).

According to incomplete statistics from the “China Artificial Intelligence Large Model Map Research Report” released by China News Network at the end of May 2023, 79 large models with over 1 billion parameters have been published in China. Here we must emphasize that although companies like OpenAI that have released large language models have paved the way for others, developing large models is a complex project with a high threshold that not every company needs to undertake. The development of large models involves large model algorithms, high-quality datasets, and powerful computing power. According to Nvidia, training GPT-3 with 8 V100 GPUs is expected to take 36 years, while using 1024 80GB A100 GPUs can shorten the training time to 1 month. Once large models are commercialized, their operational costs will be extremely high. Especially when large models are accessed frequently, computing power may become a bottleneck, potentially requiring dozens or even hundreds of times more computing power, with chip costs for enterprises potentially reaching billions of dollars. For the vast majority of enterprises, it is neither feasible nor necessary to develop their own large models; they should focus on reorganizing their core businesses and studying how large models or the large models developed in vertical fields can aid their operations. As the barriers to using large models continue to decrease, enterprises should focus on their business and customers, clarifying that technology itself does not create value but is a tool for creating value for customers. Having powerful computing power and vast amounts of data is not sufficient to ensure that a company remains at the forefront of AI technology; innovation is key, and creativity is the result of long-term cultivation of multiple factors, not something that can be simply acquired through introduction or money.

While AI technology is indeed powerful, we must have a thorough understanding of its limitations and potential risks. Large language models like AIGC have already demonstrated their outstanding capabilities in text generation tasks, including article writing, poetry creation, and programming code generation. However, these models still have certain limitations, and corporate managers must understand and carefully consider how to address their potential risks when applying them.

First, the limitations of understanding and reasoning.While these models can generate fluent, grammatically correct, and even creative texts, they do not truly understand the content they produce. Therefore, they may generate factually incorrect information or perform poorly on tasks that require deep reasoning.

Second, ethical and moral issues.Generative artificial intelligence may produce biased, untrue, or harmful content. This is because these models are often trained on vast amounts of text from the internet, which may contain biased or erroneous information.

Third, lack of transparency and interpretability.The decision-making processes of large models are often a “black box,” making it difficult to understand why a model generates specific outputs. This poses a significant challenge in fields requiring transparency and interpretability (such as healthcare and law).

Fourth, data and computing resource demands.Training these large models requires vast amounts of data and computing resources, which may limit their application in some environments.

For corporate managers, AI is both an opportunity and a challenge. Historically, whenever we face the emergence of disruptive technologies, there is concern that technology will replace humans on a large scale. Unlike many technologies that emerged during the industrialization era, AI will ruthlessly replace some human (even high-paying) intellectual labor. However, we must recognize that the real competitors are not AI itself, but those who can better leverage AI (Daugherty et al., 2023). AI will continue to change the world, and our task is to find our place in this change, seize the opportunities brought by it, and build competitive advantages that adapt to the future intelligent era. Therefore, we not only need to accept AI but also learn how to dance with it, so as to maintain a leading position in this technological revolution (Zao-Sanders and Ramos, 2023).

In this rapidly changing era, the role of government is crucial. Recently, Geoffrey Hinton, known as the “father of artificial intelligence,” reminded us that “we should not scale up AI further until we figure out if we can control it.” How to regulate and supervise artificial intelligence has become a pressing issue for governments around the world to consider (Heikkiläarchive, 2023; Beijing Municipal Government, 2023). In the context of China, while strengthening regulation, the government also has many important roles: how to assist the market in promoting the construction of computing power infrastructure? How to encourage AIGC and the subsequent AGI technologies and promote the application of AI innovation scenarios? Finding a reasonable dynamic balance between regulatory norms and encouraging innovative applications is key to the development of artificial intelligence in China.

References

[1] “Several Measures to Promote the Innovation and Development of General Artificial Intelligence,” Beijing Municipal Government Website, May 30, 2023, https://www.beijing.gov.cn/zhengce/gfxwj/202305/t20230530_3116869.html.

[2] Marco Ian Siti, Karim Lakani: “Corporate Competitive Strategy in the AI Era,” translated by Liu Zhengzheng, published in Harvard Business Review, Issue 1, 2020.

[3] Talon Karna: “When Technology Outpaces Society,” translated by Liu Xiaowei, published in Harvard Business Review, Issue 8, 2018.

[4] Wu Jihong: “AI Changes the Rules of Competition,” published in Harvard Business Review, Issue 9, 2018.

[5] Daugherty R P, Wilson H J, Narain K. Generative AI Will Enhance—Not Erase—Customer Service Jobs [J/OL]. Harvard Business Review, 2023(30), https://hbr.org/2023/03/generative-ai-will-enhance-noterase-customer-service-jobs.

[6] Eulerich M, Sanatizadeh A, Vakilzadeh H, Wood D A. Can Artificial Intelligence Pass Accounting Certification Exams? ChatGPT: CPA, CMA, CIA, and EA? [R]. SSRN Working Paper, 2023(18).

[7] Heikkiläarchive M. Suddenly, everyone wants to talk about how to regulate AI [J/OL]. MIT Tech Review, 2023(23), https://www.technologyreview.com/2023/05/23/1073526/suddenly-everyone-wants-to-talk-about-how-to-regulate-ai/.

[8] Introducing the Microsoft 365 Copilot Early Access Program and new capabilities in Copilot, 2023(9), https://www.microsoft.com/en-us/microsoft-365/blog/2023/05/09/introducing-the-microsoft-365-copilot-early-access-program-and-new-capabilities-in-copilot/.

[9] Schrage M, Kiron D, Candelon F, Khodabandeh S, Chu M. AI is Helping Companies Redefine, Not Just Improve, Performance [J]. MIT Sloan Management Review and BCG, 2023(8).

[10] Siegel E. How Machine Learning Can Improve the Customer Experience [J/OL]. Harvard Business Review, 2023(24), https://hbr.org/2023/03/how-machine-learning-can-improve-the-customer-experience.

[11] Stouthuysen K. How Digitally Mature Is Your Finance Office? [J/OL]. MIT Sloan Management Review, 2023(17), https://sloanreview.mit.edu/searchresults/cx=015845200821302873875%3Afugcafzom4o&amp.

[12] Zao-Sanders M, Ramos M. A Framework for Picking the Right Generative AI Project [J/OL]. Harvard Business Review, 2023(29), https://hbr.org/2023/03/a-framework-for-picking-the-right-generative-ai-project.