A Review of AI and Its Impact on Management Accounting and Society

Abstract

Past and current advances in artificial intelligence (AI) have resulted in a significant impact on business and accounting. Over time, AI has slowly transformed from the 1950s to today, from rule-based systems, also known as expert systems, to the deep learning architectures and sophisticated neural networks of modern generative AI. Early AI accounting applications of expert systems included a GAAP-based expert system to assess the appropriate accounting treatment for business combinations and an expert system to determine the proper type of audit report to issue. Recent accounting expert systems have been developed for document analysis, fraud detection, evaluating credit risk, and corporate default forecasting. The purpose of this study is to examine key events in the history of AI, current applications, and potential future effects pertaining to management accounting and society overall. In addition, the relationship of AI with economic and social factors will be evaluated. The study’s findings will be of interest to management accountants, businesspersons, academic researchers, and others who are concerned with artificial intelligence and its impact on management accounting and society overall.

1. Introduction

The concept of artificial intelligence (AI) goes back many years. The earliest applications of AI can be traced back to the 1950s. The term “artificial intelligence” was first used at a Dartmouth seminar in 1956. The recent expansion of AI applications in the 2020s has garnered worldwide interest. Developments in artificial intelligence (AI) have resulted in a significant impact on business and accounting. Recent times have experienced a tsunami of potential uses of AI in business and other parts of society. AI-driven robots are expected to eliminate numerous jobs, but these job losses may be offset by new jobs in other areas (Badet, 2021).

AI has developed from rule-based systems, also known as expert systems, into the deep learning architectures and sophisticated neural networks of modern generative AI, which feature vibrant, learning-based methodologies (Saxena et al., 2024). There are AI applications that enable anyone to generate images, texts, computer codes, and videos, while also offering insights that are difficult to detect using traditional tools (He et al., 2025).

Among the first accounting applications of AI were expert systems, such as the GAAP-based expert system to assess the appropriate accounting treatment for business combinations (L. M. Smith & McDuffie, 1996; R. McDuffie et al., 1991a, 1991b). Another early AI application was an expert system that determines the proper type of audit report to issue (R. S. McDuffie & Smith, 2006). Recent accounting expert systems have been developed for document analysis (Mahadevkar et al., 2024), fraud detection (A. T. Khan et al., 2022), credit risk evaluation (Machado & Karray, 2022), and corporate default forecasting (Moscatelli et al., 2020).

Using AI, management accountants can become more efficient and effective in their work. Research on AI use indicates that people using AI are able to finish tasks more efficiently and that their work will be of a higher quality. AI has the potential to bridge the ability gap separating lower-skilled and higher-skilled workers, but with the caveat that failure to provide appropriate oversight could result in lower-quality outputs (Maslej et al., 2024). Using AI can aid management accountants in a variety of tasks, such as budgeting and costing (Lee et al., 2024), maintaining and controlling inventory records (Singh & Adhikari, 2023), and managing receivables and credit approval (e.g., Machado & Karray, 2022). Possible future AI applications might assist management accountants in strategic financial planning (Yeo et al., 2024), enhancing cyber security (cf., Umoga et al., 2024), and maintaining ethical standards (cf., Ariail et al., 2024b).

The purpose of this study is to examine pivotal events in the history of AI, describe some current AI applications, and consider possible future effects on management accounting and society overall. In addition, the relationship of AI with economic and social factors will be evaluated. The study’s findings will be of interest to management accountants, businesspersons, and academic researchers who have an interest in artificial intelligence and how it has affected and will affect management accounting and society in the future.

Based on the above, two research questions are examined, as follows:

RQ1: 

What are some key events pertaining to AI, its history, current applications, and potential future effects on management accounting and society overall?

RQ2: 

What is the relationship of current AI to economic and social variables?

2. Prior Research on Information Technology Related to Economic and Accounting Issues

There have been a number of important information technology breakthroughs in history, which generally have an economic impact, with most having some impact on the field of accounting, notably management accounting and financial reporting. The earliest information technology breakthrough was the usage of clay tablets for financial record-keeping by Mesopotamian scribes 6000 years ago, regarded as the forerunners of modern-day accountants (cf., Axtell et al., 2017; Macve, 2015). The invention of the movable type printing press by Johan Gutenberg in 1455, arguably the greatest information technology innovation in history, had a revolutionary impact on the availability of books, including those for financial record-keeping (Axtell et al., 2017).

Less than 40 years after the invention of the printing press, the first printed book on double-entry accounting was authored by the Father of Accounting, Friar Luca Pacioli in 1494 (L. M. Smith, 2022; Axtell et al., 2017). Pacioli’s book could be used to teach basic accounting. Further, the book offered guidance on other accounting topics, including cost accounting and accounting ethics (L. M. Smith, 2022). Subsequent advances in science and technology paved the way for a variety of breakthroughs, including work by Sir Francis Bacon (Davis, 2019) and Charles Babbage (Miller, 1990). These early pioneers laid the foundation for new technologies, ultimately leading to the development of modern-day AI.

The first industrial revolution, which took place in the 18th and 19th centuries, included the invention of the steam engine, which transformed rural, agricultural societies into industrial and urban societies. Between 1870 and 1914, the second industrial revolution used electricity to power machines and create mass production. Commencing in the 1980s, the third industrial revolution, also known as the digital revolution, changed production functions dramatically with the invention of computers, the Internet, and information and communication technology. While there is no doubt that the information technology (IT) revolution has enhanced our quality of life, it also has had a major impact on management accounting (cf., Astuti & Augustine, 2022; Bruggeman & Slagmulder, 1995), financial reporting (cf., Grant et al., 2008), and economic growth (Imamov & Semenikhina, 2021).

A stream of literature focuses on the impact of IT on productivity, an important measure and determinant of economic performance. Earlier studies found no connection between IT investment and productivity at the firm level, industry level, or the economy as a whole (e.g., Roach, 1987; Strassmann, 1990). The lack of results led researchers to conduct more rigorous analyses of the relationship between IT and productivity with larger data samples and more refined research methods (e.g., Brynjolfsson, 1993, 1996; Bresnahan, 1999; Brynjolfsson & Hitt, 1995, 1996; Oliner & Sichel, 2000; Jorgenson, 2001).

More recent studies have found that IT investments contribute to firm productivity (e.g., Brynjolfsson, 1993, 1996; Bresnahan, 1999; Brynjolfsson & Hitt, 1995, 1996). Brynjolfsson and Hitt (2000, 2003) show that IT has a substantial impact on labor productivity. Gilchrist et al. (2001) report the same findings in the manufacturing industry. Furthermore, Atrostic and Nguyen (2005) documents that the use of computer networks increases productivity in manufacturing plants by about 7.5 percent.

In addition to research on U.S. firms, some other studies focus on firms in other countries. On one hand, this line of work supports findings based on U.S. firms. For instance, Farooqui’s (2005) research on firms in the UK indicates that a 10 percent increase in the number of workers using computers leads to an increase of 2.2 percent in productivity for older firms and 4.4 percent for newer firms. Criscuolo and Waldron (2003), in their UK-based study, report that computer network usage increases total factor productivity by 5 percent. Moreover, Mairesse et al. (2001), who focus on French firms’ IT investment and productivity, report findings consistent with prior research. On the other hand, Lal (2001) does not find a relationship between IT investment and productivity in the Indian market. Dewan and Kraemer (2000) and Pohjola (2000) find that IT plays a significant role in productivity in developed countries, but not in developing countries. Dedrick et al. (2003) argue that the low unit cost of labor and relatively high cost of capital make it difficult for capital-labor substitution in developing countries.

While various studies provide evidence of a positive association between IT investment and productivity, the impact varies among different firms (Brynjolfsson & Hitt, 1995, 1996). Black and Lynch (2001) report that a higher proportion of non-managerial employees using computers was associated with greater productivity. They argue that productivity is not affected by the presence or absence of a particular management practice, such as total quality management, but by the way in which the practice is implemented. Idiosyncratic firm characteristics and specific features of organizational structures and management practices explain part of the variation (Dedrick et al., 2003).

Consistent with firm-level analyses, studies using aggregate data at the industry level find similar results. Daveri (2003) and Dumagan et al. (2003) find that acceleration in productivity occurs in sectors that invest more in IT. Furthermore, IT investment is the major driver of national productivity growth. While there continues to be a debate over the magnitude of the IT contribution to productivity, various studies support the notion of significant and positive long-term impacts from IT investments on national productivity (Jorgenson, 2001; Oliner & Sichel, 2000; Bosworth & Triplett, 2000). The significant impact of IT on the productivity of firms applies in many other nations, including Australia (Simon & Wardrop, 2002), France (Mairesse et al., 2001), Korea (Seo & Lee, 2006), Japan (Motohashi, 2007), and Switzerland (Simon & Wardrop, 2002).

3. Prior Research and Theory Related to AI and Expert Systems

Early artificial intelligence (AI) systems began with relatively simple expert systems. AI systems advanced over the last several decades to systems that include complex algorithms and self-learning abilities, which perform similarly to the human brain. Unlike traditional machines, which replaced human labor for simple manual work and heavy or dangerous activities, AI has the capacity to do the cognitive as well as physical work of a large fraction of the labor force. As a result, there is a growing perception that advances in AI will radically transform the workplace and change the type of human work in a more comprehensive way than traditional technologies did (Brynjolfsson & McAfee, 2014; Ford, 2015; C. H. Lu, 2021).

The theoretical foundation of AI is based on information systems theory. AI systems include the same four components as all systems, specifically, input, process, output, and feedback (cf., Hevner & Chatterjee, 2010; Gregor, 2006). Thus, information systems theory lays the foundation for describing the functions of AI systems. As presented in Figure 1, information systems theory comprises four parts: input, process, output, and feedback (cf., Gregor, 2006; L. M. Smith et al., 2003). While information systems can be relatively simple or complex, as in the case of AI systems, they all have the same four fundamental components.

Figure 1. Information systems theory components. Adapted from the following: (Gregor, 2006; L. M. Smith et al., 2003).

The development and growth of artificial intelligence applications can be understood by the theory of capitalism (cf., Hong, 2024; Paganelli, 2024; Yolles, 2024). “Capitalism—unlike socialism, Marxism or authoritarianism—rewards individual innovation and work ethic. Anyone can improve their station in life by hard work or coming up with a new idea. That has rarely been the case throughout human history and is still not true in many countries in the world” (Stewart, 2020). The growth in AI applications for accounting, business, science, search activities, report-writing, visual design, and other areas has resulted from extensive consumer demand and the response of innovative companies to build and supply AI applications (Rashid & Kausik, 2024).

A number of studies have examined the impact of AI technology on the labor market from different perspectives. Consistent with the notion of workforce transformation, Morikawa (2017) surveyed over 300 Japanese firms in the manufacturing and service industries and found that firms place a priority on upgrading human capital with the mindset that the pattern of employment will change in the future. Berg et al. (2018) analyze how automation may transform the labor market with a range of variants. The results suggest that automation has a positive impact on growth, but not on equality. Real wages fall in the short run, but will rise in the long run, which may take generations. Magoutas et al. (2024) emphasize the expanding influence of AI technologies on businesses and the need to boost human capital to support structural transformation.

An emerging stream of literature focuses on understanding the impact of AI on labor productivity (e.g., Alderucci et al., 2020; Damioli et al., 2021). The European Commission (2015) reports that the use of industrial robots drives labor productivity based on a sample of manufacturing firms in seven European countries. Similarly, Damioli et al. (2021) reveal a positive impact of AI patent applications on firm labor productivity. Concentrating on small and medium enterprises and services industries, their results suggest that the ability to fully exploit the benefit of AI depends on the pace of firms to readjust and introduce AI in the production process. Kopka and Fornahl (2024) examined firms in 28 European Union member states and find that smaller frontier firms have higher labor productivity growth, but not for larger latecomer firms. The results provide implications that specific AI types may explain the diverse effects on productivity.

One of the most frequently asked questions about AI is its impact on employment (Y. Lu & Zhou, 2021). The concern of AI eliminating jobs and replacing workers is a concern to technologists, business elites, and policymakers (Berg et al., 2018; Y. Lu & Zhou, 2021). Consistent with that notion, Acemoglu and Restrepo (2020) document a negative association between the use of robotics and employment, indicating that the use of one robot per thousand workers decreases employment to population ratio by 0.18–0.34 percent. Furthermore, Frey and Osborne (2017) derive an index to investigate the susceptibility of occupations to automation. Their study shows that employment in the service sector tends to be susceptible to automation in the United States. Similarly, Brynjolfsson and McAfee (2014) and Ford (2015) provide evidence that advances in AI technology leads to unemployment. A 2023 McKinsey survey, which questioned business professionals regarding expectations about AI’s effect on workforces in upcoming three years, found that 30 percent anticipated small or zero changes in staff size, while 43 percent felt staff would shrink (Maslej et al., 2024).

In contrast to the above research indicating that AI replaces existing jobs, Graetz and Michaels (2018) find no significant effect from industrial robots on overall employment. The study provides the first analysis of the economic impact of industrial robots using data from a panel of industries in 17 countries. Although the working hours of low-skilled and middle-skilled workers are reduced, the result shows no significant impact from industrial robots on overall working hours.

Recognized as one of the most important technological innovations in the 21st century, AI is booming and is capable of transforming every aspect of our social interactions from the government to businesses, industries, healthcare, and education. Artificial intelligence in education (AIEd) has received a lot of attention in recent years and continues to gain momentum. A number of studies indicate that students, teachers, and administrators all benefit from an education system with a greater exposure to AI-based technologies in classrooms (e.g., Benvenuti et al., 2023; Grover et al., 2015).

By tailoring the individual needs and learning styles of students, AI delivers a personalized educational experience that surpasses the limitations of conventional classroom experiences (Aleven et al., 2016). Benvenuti et al. (2023) show that AI promotes creativity, critical thinking, and problem-solving in educational contexts. The personalized approach allows students to take control of their own learning and facilitates collaborative learning and peer interactions, creating a more dynamic and engaging learning environment (Grover et al., 2015). In addition, the integration of AI enables educators to devote more time to the complex and creative aspects of teaching. AI offers educators the ability to identify students’ learning needs and address them promptly through customized assessment and automated grading (Celik et al., 2022; Tuomi, 2018).

According to Pedro et al. (2019), one of the major challenges of AI application lies in ensuring inclusion and equity. Despite the benefits, the adoption of AI in classrooms may deepen the existing inequalities as the marginalized and disadvantaged population are more likely to be excluded from AI-powered education (cf., Hilbert, 2016; Zawacki-Richter et al., 2019). Welham (2008) argues that a lack of money or time is the main reason that prevents public institutions from developing and implementing AI-based technologies. Moreover, Nye (2015) identified seven barriers for introducing AI in education in developing countries, including hardware availability, electrical availability, Internet reliability, data costs, students’ basic internet skills, language, and a lack of culturally appropriate content. To remove these obstacles, policies would need to be established that define the use of the internet as a right and create international alliances to build infrastructures in poorer and resource-constrained developing contexts (Pedro et al., 2019; Mutoni, 2017).

From a historical perspective, past to present, Table 1 lists some of the key events in the history of AI. First on the list, as previously mentioned, is Sir Francis Bacon, who is credited with opening the way for modern science, notably, promoting the scientific method and highlighting observation and experimentation, rather than relying on old-style Aristotelian reasoning (cf., Quinton et al., 2025; World History Encyclopedia, 2025; Davis, 2019). Some have suggested that Bacon would support the development of AI, while at the same time being concerned and cautionary.

Table 1. Notable events for AI.

Year	Event
1605	Francis Bacon authors “The Advancement of Learning,” advancing the scientific method.
1812	English polymath Charles Babbage conceives the idea of an automated digital computer.
1926	A German movie, “Metropolis,” depicts AI in the robot character, Maria, introducing people to sentient, human-like machines.
1950	Alan Turing, an English computer scientist and mathematician, publishes a paper titled “Computing Machinery and Intelligence,” which lays out what has become known as the Turing Test to determine whether a machine is capable of thinking.
1956	The term “Artificial Intelligence” is first introduced by John McCarthy, a mathematics professor at the Dartmouth workshop. The workshop marked the formal inception of AI as an academic discipline and is widely considered as the birthplace of the new field.
1968	In the American movie, “2001: A Space Odyssey,” a spacecraft is controlled by an AI supercomputer named HAL 9000.
1979	The American Association of Artificial Intelligence is formed. The association later evolved into the Association for the Advancement of Artificial Intelligence (AAAI).
1986	A Mercedes van, the first self-driving car, is built in Munich under the direction of scientist Ernst Dickmanns. Equipped with a computer system and sensors, the vehicle could read the environment and drive up to 55 mph on empty roads.
1991	Early accounting applications of expert systems include a GAAP-based expert system, PURPOOL, to assess the proper accounting treatment for business combinations.
1997	Deep Blue, a computer system developed by IBM, wins the rematch with the world chess champion, Gary Kasparov. As the first program to beat a human chess champion, Deep Blue could review 200 million potential chess moves in one second.
2001	Steven Spielberg writes the screenplay and directs “A.I. Artificial Intelligence,” an acclaimed 2001 American movie about an 11-year-old AI robotic child.
2001	Microsoft Game Studios releases a video game, Halo: Combat Evolved, in which players take the role of Master Chief Petty Officer John-117, who relies on an AI character, Cortana, for help.
2006	An expert system, AUDPORT, is constructed to help students understand various audit reporting issues and corresponding report formats.
2011	Apple introduces the first popular virtual assistant, named Siri. Amazon releases its own virtual assistant named Alexa three years later. With natural language processing capabilities, Siri and Alexa are programed to understand and answer a lengthy list of questions.
2012	Geoffrey Hinton and his graduate students present their research on neural networks at the ImageNet challenge. Built to process data in a manner similar to the human brain, neural networks are a breakthrough in the field of computer vision.
2016	AlphaGo, developed by Google DeepMind, defeats the ancient game Go champion Lee Sedol. AlphaGo combines neural networks and advanced search algorithms to play the game using a method called reinforcement learning.
2016	Hong Kong-based company Hanson Robotics creates the first humanoid robot, named Sophia, with a human appearance and the ability to communicate and replicate emotions.
2022	OpenAI releases the AI chatbot ChatGPT to the public. The interface surpassed one million users in just one week after its release, becoming the fastest-growing consumer application in history.
2024	AI chatbot ChatGPT is the 9th most downloaded app worldwide.

Adapted from L. M. Smith and Xu (2025). Notable events in the history of AI. Available at the following link: http://bit.ly/notable-events-AI (accessed on 14 June 2025). Used with permission.

Second on the list is Charles Babbage, mentioned earlier, who originated the concept of an automated digital computer (cf., Freiberger & Swaine, 2025; Miller, 1990; Halacy, 1970) and thereby became known as the father of the computer. Without the foundation provided by Bacon and Babbage, it is doubtful scientific progress would have achieved AI in the modern day (cf., Andrews, 2025; Davis, 2019). Next mentioned is the presentation of artificial intelligence in a fictional character, Maria, a robot, in the 1926 German movie, Metropolis, a ground-breaking concept at the time (Wosk, 2010). Other notable events include Alan Turing’s work on determining whether a machine is actually able to think (Saha et al., 2024), coining of the term “artificial intelligence” (Pasham, 2024), and OpenAI’s release of the AI chatbot ChatGPT, which became the fastest-growing consumer application of all time (Gordon, 2023).

4. Examples of AI and Expert System Applications and Effects on Management Accounting

This section provides examples of AI and expert system applications that have been developed in recent years. These applications have been used in accounting and business, including use by management accountants. The following types of applications are reviewed below: voice assistants, chatbots, and expert systems.

4.1. Voice Assistants

In recent years, the use of AI has become ubiquitous and has made many aspects of people’s lives easier. One use of AI that is particularly popular today that can be used by management accountants is voice assistants that use spoken language to execute commands on devices such as smart TVs, smart speakers, computers, and smartphones. These voice assistants include Alexa (by Amazon), Siri (by Apple), Cortana (by Microsoft), Bixby (by Samsung), and Gemini Live (by Google). AI voice technology makes it possible for people to talk naturally to these devices, improving how people interact with technology, how customers interact with companies that utilize this technology, and how company employees perform their jobs. At the end of 2024, there were approximately 146 million users of voice assistants in the United States; that number is expected to increase to 157.1 million by 2026 (Lan et al., 2024; Tsymbal, 2024).

In a 2023 survey of 400 business leaders in the United States and Canada, Deepgram found that 82 percent of companies were using voice technology, an increase of 6 percent from the year before. In companies with more than 2500 employees, the 2023 percentage jumped to 92 percent (Deepgram, 2023). Among the benefits reported by company employees who use AI voice technology in their jobs are the following: (1) ease of use, as speaking is more natural than typing, (2) hands-free operation, allowing employees to perform tasks with both hands, and (3) speed, as speaking is faster and easier than keyboarding. When used to interface with customers/clients, additional company benefits of AI voice technology include increased profitability through increased sales and decreased costs of customer service. Using AI voice technology also improved inclusiveness and accessibility for customers with physical disabilities that limit the use of text-based interfaces (Tsymbal, 2024).

4.2. Chatbots

Another popular use of AI involves generative AI chatbot applications and their underlying Large Language Model (LLM) technology. Chatbots allow management accountants to interact with computers and other devices using natural human dialog. Accountants can ask questions and receive responses that are nearly identical to responses that humans would produce. This ability comes from a chatbot’s use of an LLM consisting of large sets of data used to train the model to predict appropriate responses. A simple description of this training is provided by Hannigan et al. (2024):

“A generative chatbot uses neural networks to learn the semantic distance between words using vector coordinates assigned to the words. For instance, with English language training data, the word ‘pasta’ follows the word ‘eat’ more often than the word ‘chaos,’ and in a vector space of words ‘pasta’ and eat’ would be closer to each other than ‘chaos.’ This is how chatbots generate new human-like text-based responses to the prompts they receive.”

(Hannigan et al., 2024, p. 5)

While LLMs function by predicting responses, they are not (yet) capable of understanding the meaning of their responses, which led Bender et al. (2021) to describe Chatbots and their underlying LLMs as “stochastic parrots.”

“Contrary to how it may seem when we observe its output, an LM [language model] is a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning: a stochastic parrot.”

(Bender et al., 2021, p. 617)

In the following paragraphs and Table 2, we describe some popular Chatbots currently in use.

Table 2. Popular Chatbots.

Name	Company	Release Date	Availability	Large Language Model
ChatGPT	Open AI	November 2022	Free at https://chatgpt.com	GPT LLM
Meta AI	Meta	February 2023	Free at https://www.meta.ai	Llama LLM
Claude	Anthropic	March 2023	Free at https://claude.ai and via Android and iOS apps.	Claude LLM
Copilot	Microsoft	November 2023	Free at https://copilot.microsoft.com and via Android and iOS apps. Included with Windows 11, Office 365, and Copilot+ PCs.	GPT LLM
Gemini	Google	December 2023	Free at https://gemini.google.com/app and via Android and iOS apps.	Gemini LLM
Grok AI	xAI	December 2023	Free at https://x.com and via X (formerly Twitter) on iOS and Android.	Grok LLM
DeepSeek	High-Flyer	January 2025	Free at https://chat.deepseek.com and via Android and iOS apps.	DeepSeek-LLM

One widely used generative AI chatbot is ChatGPT. ChatGPT is based on the GPT LLM (Generative Pre-trained Transformer Large Language Model) developed by the OpenAI company and was released in November 2022. Within two months of its release by OpenAI, ChatGPT had approximately 100,000 users. Within one year, it was being used by more than 90 percent of Fortune 500 companies. A free version of ChatGPT can be accessed at https://chatgpt.com (Ballantine et al., 2024; Deepgram, 2023). In 2024, ChatGPT became the 9th most downloaded app worldwide (Larson, 2025) Soon after the release of OpenAI’s ChatGPT, several other companies released their own chatbots. In February 2023, we saw the launch of Microsoft Bing Chat and Meta AI. Bing Chat was subsequently renamed Microsoft Copilot in November 2023 and uses OpenAI’s GPT LLM. Copilot is integrated into Office 365 applications, including Word, Excel, PowerPoint, Outlook, and Teams. Copilot is also integrated into the Windows 11 operating system and Copilot+ PCs, which are laptop computers designed specifically for AI that include a neural processing unit (Ortiz, 2024; Whitney, 2025).

Meta AI, also released in February 2023, is a generative AI chatbot based on Meta’s Llama LLM. The latest releases are multilingual and can respond to user requests in various languages, including French, German, Hindi, Italian, Portugues, and Spanish. Unlike OpenAI’s GPT LLM used by ChatGPT and Copilot, Llama LLM is open source and can be downloaded and used free of charge. Meta AI can be accessed at www.meta.ai, as well as on Facebook, Messenger, WhatsApp, and Instagram (AGI Sphere, 2023; Fried, 2024; Softweb Solutions, 2024).

In response to ChatGPT’s success and a perceived threat to Google Search, Google released its Bard generative AI chatbot in March 2023. An updated and improved version was launched in December 2023 renamed as Gemini, which uses an LLM of the same name. Gemini has been found to perform on par or better than OpenAI’s GPT-4 in standard benchmark tests. Gemini is available to use free at gemini.google.com, on Google’s Pixel smartphones, and on Android and iOS devices via the Gemini app. A subscription version, Gemini Advanced, provides improved performance and additional features like Gemini Live, which allows users to converse back and forth with AI without typing, and is integrated into Google’s Gmail, Docs, Sheets, and Maps (Akhtar, 2024; Hawkins, 2024; Heaven, 2024).

Anthropic Corporation was founded in 2021 by a group of research scientists and engineers who were former employees of OpenAI. Anthropic’s purpose and mission include generating research about the opportunities and risks of AI and the development of AI systems that help people and society flourish and are aligned with human values, safety, transparency, and ethical principles (Anthropic, 2025).

In March 2023, Anthropic released the first version of Claude, Anthropic’s generative AI chatbot. Unlike other chatbots, each version of Claude’s LLM is closed source and unable to update itself by retrieving data from the Internet, relying only on the large dataset on which that version of the LLM was trained. Nevertheless, Claude has been heralded as a significant advance in AI chatbots by using a “constitutional AI” framework that increases the likelihood its output will be honest, unbiased, and harmless and reduces the possibility the output can be used for illegal or unethical purposes. The editors at CNET awarded Claude the title of “Best Overall Chatbot of 2024” based on its constitutional framework, the high quality of its responses, and its ability to ask clarifying questions as well as change its tone and depth of responses based on the user’s input. Claude is accessible at https://claude.ai and via the Claude apps for iOS and Android (Anthropic, 2023; I. Khan, 2024; Pazur, 2025).

The Grok AI chatbot was created by xAI, a company devoted to using AI to “accelerate human scientific discovery.” Grok was first released in December 2023 to X users (formerly Twitter) who subscribed to X Premium+. Version Grok-2 was released a year later in December 2024 and is available free to all users of X on iOS, Android, and on the web at https://x.com. Grok has the unique distinction of being described as a “humorous AI Assistant.” Grok’s LLM is unique in that its dataset includes posts from X as well as other websites across the Internet. Grok’s responses also include citations, which help users to verify the accuracy of the information in Grok’s responses (Roth, 2024; X Help Center, n.d.; X AI Blog, 2024).

As of this writing, the latest chatbot released to the public was DeepSeek, which was launched in January 2025 by High-Flyer, a Chinese hedge fund. Initial tests indicate DeepSeek’s performance and user interface compare favorably with OpenAI’s ChatGPT. However, DeepSeek’s LLM’s development and usage costs are reportedly 20 to 50 times lower than other LLMs due to better energy efficiency of its technology. In addition, DeepSeek’s code is open-source and is available to the public free of charge at https://chat.deepseek.com and via Android and iOS apps on smart devices (Baptista, 2025; Browne & Butts, 2025; Forlini, 2025; Ng et al., 2025).

4.3. Expert Systems

The general objective of AI rules-based systems, that is, expert systems, is to replicate the decisions of human experts. This is accomplished by the AI system going through a series of if–then rules or questions. An expert system operates by using a knowledge base of domain-specific facts, a set of rules related to the domain, and an inference engine that uses the rules to analyze the facts and make predictions or recommendations. Unlike chatbots and voice assistants, expert systems are designed for use in relatively narrow fields requiring a high degree of knowledge (i.e., “expertise”) like medicine, law, and the sciences. Although expert systems have existed much longer than chatbots, their use is currently much less widespread. Dating back to the 1960s and 1970s, DENDRAL and MYCIN were two of the first expert systems. DENDRAL was designed by a team of computer scientists at Standford University to mimic experienced organic chemists’ decision-making processes when attempting to identify organic molecules. One member of the team, Ed Feigenbaum, once remarked, “we were trying to invent AI, and in the process discovered an expert system” (Lindsay et al., 1993; Press, 2022). Soon thereafter, the team went on to develop the MYCIN expert system to help medical practitioners diagnose blood infections. MYCIN’s diagnostic accuracy was reported to be on par with human blood infection specialists and somewhat better than general medical practitioners (Copeland, n.d.).

Since the beginning of the 21st century, interest in expert systems has declined significantly compared to other AI applications, when measured in terms of Google search trends (Preis, 2023). Nevertheless, researchers have recently developed, and are still developing, new systems to facilitate expert decision making for a wide variety of purposes, including estimating the level of risk of fatal ventricular arrhythmia (Galán et al., 2024), clinical decision support for viral infections and outbreaks (Chrimes, 2023), prevention of social engineering attacks (Cardoso et al., 2023), cyber security related to the Internet of Things (IoT) (Rak et al., 2022), complex project management (Bhattacharya et al., 2021), identification of pathogens and diseases (Abu Al-Qumboz & Abu-Naser, 2019; Elsharif & Abu-Naser, 2019), marketing (Papadopoulos et al., 2018), and many others.

While overall expert system use may have declined, the potential for expert system use in management accounting is quite high. Example uses of expert systems in management accounting that have already been developed or could be developed include budgeting and financial planning, inventory management and control, standard costing and anomaly analysis, financial reporting evaluation, customer credit analysis, and fraud detection.

5. Anticipated Future Developments Regarding AI and Their Economic Impact

The use of AI by companies and consumers has affected business in many parts of the world and will continue to spur growth in commerce and the global economy. By 2030, the cumulative effect of AI since 2017 on global GDP is expected to reach more than USD 15 trillion from the use of AI to enhance products and services and stimulate consumer demand in industries such as automotive, financial services, transportation, retail, healthcare, and energy (Harari, 2024, p. 374; Rao & Verweij, 2017). In these industries, management accountants will be using AI or working with AI users in their organizations.

AI specialists at PricewaterhouseCoopers anticipate that 55 percent of all AI-related GDP gains will arise from improvements in productivity from the automation of repetitive tasks and the expansion of employees’ capabilities primarily in manufacturing and transportation. As AI leads to the improvement and expansion of products and services, consumer demand is expected to increase, thereby driving further increases in global GDP. Sectors of the global economy expected to experience the most growth from product and service enhancements due to AI include the healthcare sector, the automotive sector (i.e., original equipment parts, repair and aftermarket parts, financing, and personal mobility as a service), and the financial services sector (i.e., banking, insurance, and wealth management). The economies of China and North America are expected to benefit the most from AI, with projected GDP gains of 26 percent (USD 7 trillion) in China and 14.5 percent (USD 3.7 trillion) in North America (PricewaterhouseCoopers, 2025a, 2025b).

One area within AI that is expected to continue to grow is the use of AI-powered voice assistants by consumers and businesses. As mentioned previously, the number of voice assistant users in the United States is expected to increase from about 146 million in 2024 to more than 157 million by 2026. Globally, the market for voice assistants was valued at approximately USD 4.6 billion in 2022; it is predicted to grow to over USD 38 billion by 2031 (Tsymbal, 2024).

Another use of AI that is likely to continue developing and improving with time is text-to-image and text-to-video generation. Examples of text-to-image models currently available on the web are Dall-E from OpenAI, Emu from Meta, Adobe Firefly from Adobe, ImageFX from Google, Stable Diffusion from Stability AI, and Shutterstock AI Image Generator from Shutterstock. In addition, Google’s Gemini, Microsoft’s Copilot, and X’s Grok can generate images from users’ text prompts.

Text-to-video AI generators allow users to create brief video clips consisting of sequences of images by typing a description of the desired video. Relative to other applications of AI, text-to-video generators are rather immature, but their features and the quality of their output are expected to improve significantly in coming years. A few of the companies offering text-to-video AI generators to the public as of this writing include Synthesia (https://www.synthesia.io), Invideo (https://invideo.io), Hour One (https://hourone.ai), Pictory (https://pictory.ai), Kapwing (https://www.kapwing.com), and Clossyan (https://www.colossyan.com).

A third area of AI that is expected to grow in coming years is the use of AI-powered agents, also known as Agentic AI. Agentic AI can be described as follows: “Agentic AI generally refers to AI systems that possess the capacity to make autonomous decisions and take actions to achieve specific goals with limited or no direct human intervention” (PricewaterhouseCoopers, 2024).

Agentic AI surpasses Chatbots and simple automation in its ability to make decisions and complete tasks ranging in complexity from “Book me a flight from Houston to Maui” to “Our sales team needs to improve customer satisfaction rates. Analyze our current products and customer data, identify key patterns and relationships, and develop strategies for improving customer satisfaction” (Confluent, 2025). According to AI experts, Agentic AI has the potential to benefit many sectors of the economy. For instance, in the healthcare industry, Agentic AI could be used to “autonomously review claims, verify documentation, and resolve discrepancies, reducing approval times as well as manual workload while increasing customer satisfaction.” In the travel industry, Agentic AI could be used in airline operations to determine optimal pricing for flights in real time based on analysis of data about factors such as customer travel patterns and competitor pricing. When used in research and development (R&D) in the automotive and aerospace industries, AI experts believe Agentic AI could reduce time-to-market by as much as 50 percent and lower costs by 30 percent (PricewaterhouseCoopers, 2025a, 2025b).

In management accounting, Agentic AI can assist with analyzing and performing tasks involving the revenue cycle, the purchasing cycle, the cash receipts cycle, the cash disbursements cycle, and the closing and consolidation processes. Management accountants and internal auditors can also use agentic AI to perform activities such as three-way matches, analyzing and investigating anomalies in transactions and deviations from established internal control policies and procedures, continuous auditing, and tasks related to forecasting, budgeting, and control (DeLoach, 2025).

The benefits of Agentic AI are already emerging. For instance, JPMorgan Chase’s Contract Intelligence Agentic AI system is saving the company 360,000 h of manually reviewing legal documents annually. In healthcare, the Mayo Clinic is using Agentic AI in its radiology practices, which has reduced diagnostic time by 30 percent and lowered costs by reducing unnecessary procedures by 15 percent. In the transportation industry, DHL uses Agentic AI to predict shipping demands, optimize routes, and manage warehouse operations, resulting in a 15 percent reduction in operational costs and a 20 percent improvement in delivery times (PricewaterhouseCoopers, 2024). In a 2023 McKinsey report, the future economic impact of generative AI was estimated. Revenues in the high technology industry were expected to grow more than USD 240 billion, banking more than USD 200 billion, pharmaceuticals and medical products more than USD 60 billion, and education more than USD 120 billion (Maslej et al., 2024).

6. Measuring AI and Its Relationship to Economic and Social Factors: Methodology and Analysis

While the proponents of AI foresee many positive impacts, there is no universal agreement (Krishna, 2024). As previously discussed, there are many ways that AI, especially expert systems, can benefit management accounting, as well as other fields of work, leading to positive future impacts on many fields of work and, thereby, economic development and society overall. To test the relationship between current AI and economic and social factors, we examine data using the Global AI Index, GDP per capita, and the Social Progress Index. If a positive relationship exists between the Global AI Index and economic and social variables, this might foreshadow future positive impacts of AI on economic development and society overall.

The Global AI Index benchmarks nations based on the extent of innovation, investment, and implementation regarding artificial intelligence (White & Cesareo, 2025). The Index is based on 122 indicators, obtained from 24 public and private sources of data and from 83 governments. These are categorized according to three pillars and seven sub-pillars. The first pillar is Implementation, with the following sub-pillars: talent, infrastructure, and operating environment. The second pillar is Innovation, with the following sub-pillars: research and development. The third pillar is Investment, with the following sub-pillars: government strategy and commercial. The commercial variable emphasizes the extent of AI startup activity, along with the level of business initiatives and investment (White & Cesareo, 2025).

To evaluate the relationship of current AI to economic and social variables, we employ a methodology used in numerous prior studies (e.g., De Leon et al., 2024; Ariail et al., 2024a; Broker et al., 2019; K. T. Smith et al., 2019; Martin & Smith, 2015). In this methodology, in the ranking of sample countries, a top subsample is compared to a bottom subsample to identify any significant differences that might exist. For example, in the study by De Leon et al. (2024), countries were ranked by the internet usage percentage of the population. A top and bottom subsample were compared. The findings showed that internet usage had a significant negative relationship with income inequality but a significant positive relationship with life satisfaction (happiness).

The dataset for this study’s analysis includes the Global AI Index (GAI), GDP per capita, and the Social Progress Index for 69 countries, all for which complete data are available. GDP per capita is a nation’s gross domestic product divided by the nation’s population. GDP per capita is widely used in research as an indicator of a nation’s economic output on a per person basis. The Social Progress Index (SPI) is tabulated by the Social Progress Imperative. The SPI is derived from 57 indicators spanning three key dimensions: foundations of well-being, basic human needs, and opportunity (Social Progress Imperative, 2025). The SPI spans from 0 to 100, with 0 representing weak social progress and 100 representing the optimum level of social progress. The 69 countries, ranked by GAI, are shown in Table 3.

Table 3. Global AI Index, GDP per Capita, and Social Progress by country.

Rank	Country	Global AI Index Rank *	GDP per Capita (in USD) **	Social Progress Index (SPI) **
1	United States	1	65,280	85.7
2	China	2	10,217	66.1
3	Singapore	3	65,641	85.5
4	United Kingdom	4	42,354	88.5
5	France	5	40,380	88.8
6	Germany	7	46,468	90.6
7	Canada	8	46,327	91.4
8	Israel	9	43,589	83.6
9	India	10	2101	56.8
10	Japan	11	40,113	90.1
11	Switzerland	12	85,300	91.4
12	Netherlands	13	52,295	91.1
13	Saudi Arabia	14	23,140	65.1
14	Finland	15	48,712	91.9
15	Australia	17	55,057	91.3
16	Spain	18	29,565	88.7
17	Luxembourg	19	114,685	89.6
18	United Arab Emirates	20	43,103	70.6
19	Denmark	22	60,213	92.1
20	Ireland	23	80,779	90.3
21	Italy	24	33,567	87.4
22	Sweden	25	51,687	91.6
23	Norway	26	75,826	92.7
24	Belgium	27	46,414	89.5
25	Austria	28	50,122	89.5
26	Portugal	29	23,285	87.8
27	Brazil	30	8897	73.9
28	Estonia	32	23,718	87.3
29	Malta	33	30,186	84.9
30	Turkey	34	9127	68.3
31	Poland	36	15,695	84.3
32	Slovenia	37	25,941	87.7
33	Chile	38	14,742	83.3
34	Malaysia	39	11,414	77.0
35	Iceland	40	68,883	91.1
36	Hungary	41	16,733	81.0
37	Greece	42	19,151	85.8
38	Thailand	43	7817	70.7
39	Croatia	44	14,944	81.9
40	Mexico	45	9946	73.5
41	Lithuania	46	19,555	84.0
42	Argentina	47	9912	80.7
43	New Zealand	48	41,999	91.6
44	Indonesia	49	4135	69.5
45	Romania	50	12,890	78.3
46	Colombia	51	6425	74.0
47	Bulgaria	53	9828	79.9
48	Qatar	54	62,088	70.6
49	Ukraine	55	3663	73.4
50	Uruguay	56	17,688	83.0
51	Mauritius	59	11,098	79.0
52	Peru	61	7028	74.2
53	Bahrain	62	23,443	66.6
54	Jordan	63	4405	71.5
55	Oman	64	15,343	71.4
56	Armenia	65	4623	76.5
57	Philippines	67	3485	66.6
58	South Africa	69	6001	70.3
59	Latvia	70	17,794	83.2
60	Tunisia	71	3352	75.0
61	Ghana	72	2210	64.9
62	Benin	74	1220	55.6
63	Pakistan	76	1285	49.3
64	Azerbaijan	78	4806	64.1
65	Morocco	79	3230	66.9
66	Algeria	80	3976	69.9
67	Kenya	81	1817	57.1
68	Sri Lanka	82	3852	73.2
69	Ethiopia	83	856	48.6
	Mean		26,977	78.7
	Top 23 Mean		50,278	85.3
	Bottom 23 Mean		9265	69.2

Notes: * White and Cesareo (2025), ** De Leon et al. (2024).

We compare the GDP per capita and social progress of the top tercile of countries, ranked by the GAI, to the bottom tercile. The results of the comparison of the countries with the highest GAI to the countries with the lowest GAI show a significant difference in both GDP per capita and social progress. As shown in Table 4, the mean GDP per capita of the top tercile, ranked by GAI, is USD 50,278, and the mean of the bottom tercile is USD 9265. T-test findings reveal a significant difference (p < 0.000). A higher GAI is associated with higher levels of GDP per capita. Similarly, higher-GAI countries are associated with higher levels of social progress. The mean SPI is 85.3 for the top tercile, ranked by GAI, compared to a mean SPI of 69.2 for the bottom tercile. T-test findings again reveal a significant difference (p < 0.000). A higher GAI is associated with higher levels of social progress.

Table 4. T-Test Results Regarding GDP per Capita and Social Progress by Country Ranked by Global AI Index.

	GDP per Capita (USD)	Social Progress Index
Top Tercile Mean	50,278	85.3
Bottom Tercile Mean	9265	69.2
t Stat	7.1	5.6
Prob. Significance	p ≤ 0.000	p ≤ 0.000

7. Conclusions

The first research question regards key events pertaining to AI, its history, current applications, and potential future effects on management accounting and society overall. A number of key events were identified, such as Alan Turing’s work investigating whether a machine is capable of thinking; Ernst Dickmanns’s development of the first self-driving car; early accounting expert systems such as the GAAP-based expert system, PURPOOL; Apple’s introduction of the first popular virtual assistant, Siri; and OpenAI’s release of the AI chatbot ChatGPT to the public. AI expert systems in management accounting that have been used or could be used in the future include applications for inventory management and control, fraud detection, budgeting and financial planning, standard costing and anomaly analysis, financial reporting evaluation, and customer credit analysis.

The second research question concerns the relationship of current AI to economic and social variables. A comparison was made between the GDP per capita and social progress of the top tercile of countries, ranked by the Global AI Index, to the bottom tercile. The findings indicate that there is a significant difference in both GDP per capita and social progress between the higher-GAI countries and the lower-GAI countries. Among countries, higher levels of the GAI are associated with higher levels of economic activity (GDP per capita). Similarly, higher-GAI countries are associated with greater levels of social progress. This suggests that advances in AI in management accounting and other fields will likely be associated with locations having higher levels of economic activity and social progress.

The development of artificial intelligence is expected to continue in future years in management accounting, other business fields, and society at large. While many future AI applications are anticipated, there will likely be some unexpected applications as well. Not all future effects of AI are clear, such as whether AI will cause net lower employment or whether decreases in employment will be offset by increases in employment in other areas. Findings of this study should be of interest to management accountants, business people, academic researchers, and others concerned with how AI has affected and will affect business, management accounting, and society in the future.

8. Limitations and Future Research

This study was limited by the research literature examined. While the literature review was extensive, future studies could expand the literature review to include additional past research. The study was also limited by the variables used in the study. Future studies can include other economic and social variables. Finally, the study is limited by the current time period of the study. Future studies can examine additional time periods, both past and future.