How Artificial Intelligence May Impact the Accounting Profession - The CPA Journal

The rising popularity of artificial intelligence (AI) has made many CPAs feel unsure about the ways in which their profession may be affected. There are already some tedious tasks that AI can help humans complete, such as summarizing longer documents for ease of reading. As AI technology improves, many professionals are looking for other scenarios where AI would be a useful tool. By running several test prompts through ChatGPT 4.0, this article highlights the ways in which AI can assist accountants in their daily workflow, while pointing out the ways in which it falls short.

***

On November 30, 2022, OpenAI launched ChatGPT 3.5. Since then, technology companies such as Google, Meta, and Anthropic have introduced various forms of Artificial Intelligence (AI) services, making AI a prevalent topic of conversation. Human inputs, known as prompts, instruct AI models to perform certain tasks. These AI services, including ChatGPT, are referred to as generative AI because they use Large Language Models (LLM) to generate words, pictures, and videos based on prompts.

Companies implement AI in anticipation of improving productivity. According to a Thomson Reuters survey, “1 in 10 accounting and tax professionals are currently using ChatGPT/generative AI or planning to integrate these technologies into their operations” (“ChatGPT and Generative AI Within Accounting Firms and Corporate Tax Departments,” Thomson Reuters Institute, https://tinyurl.com/2ufzzcfn). The powerful automation capabilities of AI have raised concerns about its potential impact on accounting jobs.

According to surveys and interviews conducted by Thomson Reuters in May 2023, only “51% of accounting professionals consider ChatGPT and generative AI should be applied to tax, accounting or audit work” (Thomson Reuters 2023). The authors anticipate that AI will automate certain accounting jobs for efficiency, reshaping and transforming the accounting profession. Consequently, AI can potentially generate accounting job opportunities specifically in areas requiring analysis and human intervention. According to the U.S. Bureau of Labor Statistics, the occupational outlook for accountants and auditors over the next decade calls for the field to increase by nearly 6%—faster than the average growth rate for all occupations (“Occupational Outlook Handbook,” U.S. Bureau of Labor Statistics, https://tinyurl.com/b7pnrurj). The authors think this predicted growing number of accounting jobs can be attributed to roles requiring collaboration between humans and AI models. This argument is consistent with the projected 2033 employment in accounting and auditing occupations by the U.S. Bureau of Labor Statistics: about 91,000 jobs will be added, which suggests 5.8% growth from 2023 (“Occupations with the Most Job Growth,” U.S. Bureau of Labor Statistics, https://tinyurl.com/5enpj5f4).

A 2024 blog by Vision.cpa predicts, “As accounting with AI becomes more efficient and accessible, it increases in both demand and sophistication, leading to higher value and more investment in the field” (“AI Will Increase Demand for Accounting,” Vision.cpa, Mar. 31, 2024, https://tinyurl.com/3ehp65t8). AI can also assist accountants with non-repetitive tasks that have traditionally required a lot of effort to achieve, such as processing lengthy documents and extensive data sets.

This article provides examples demonstrating AI’s ability to assist accountants in tasks requiring human inputs. The authors explore two cases in which accountants can direct AI to achieve certain outcomes: (1) understanding an update in a FASB exposure draft and (2) evaluating a company’s going concern status and suggesting audit procedures. In these examples, AI cannot fully execute the job functions independently, but it can provide guidance to accountants on procedures to be pursued. They demonstrate the importance of accountants in guiding AI on achieving goals of accounting tasks while AI provides support. The examples provided support the notion that AI will not completely replace humans but instead shift or even transform the roles of accountants.

This article employs ChatGPT 4.0. This version of ChatGPT offers a more complete rationale behind its output than previous versions do. Such underlying logic helps users understand the “thought” process by which AI generates an output, thus providing information for accountants to determine the validity of the output. Having the rationale behind the output also facilitates accountants’ ability to detect instances of “hallucination,” a phenomenon whereby AI generates fabricated or unreliable information.

Understanding a FASB Exposure Draft

Staying up to date with changes in accounting standards is one of the challenges faced by accountants. Accounting standards are long, text-based documents that take time to read and comprehend. AI can help accountants stay up to date with accounting standards updates. In the following example, the authors asked ChatGPT 4.0 to analyze FASB’s exposure draft on measurement basis in the conceptual framework, issued on December 21, 2023.

To simulate comprehending the exposure draft, the authors uploaded the file to the prompt box, assigned the role of an accountant to ChatGPT 4.0, and instructed it to review the document. It is crucial to clearly instruct ChatGPT about the role it should assume in order to ensure alignment with the desired output. The prompt was written as follows (full transcript at https://tinyurl.com/2rvfwmh4):

You are an accountant, and you need to keep updated about the new accounting standards. The Financial Accounting Standards Board (FASB), which issues accounting standards in the U.S. just released a new exposure draft on the measurement chapter of Conceptual Framework for Financial Reporting. Please read the exposure draft attached. After reading it, ask me what tasks you need to do to understand the exposure draft.

After reviewing the uploaded exposure draft, ChatGPT 4.0 requested guidance:

“I’ve reviewed the exposure draft on the measurement chapter of the Conceptual Framework for Financial Reporting. Could you guide me on what specific tasks I need to perform next to deepen my understanding of the draft?”

This response from ChatGPT gave the authors reassurance that it completed reading the exposure draft as instructed and sought instructions on the next step. Subsequently, ChatGPT was given two additional tasks:

Please do the following two tasks. First, find the current accounting standards on the measurement section of Conceptual Framework for Financial Reporting. Second, summarize the difference between the current accounting standards and the exposure draft.

This task is important, because it gauges ChatGPT’s ability to comprehend the exposure draft and effectively compare it to existing GAAP. The output contains extensive content, so select excerpts are included below in quotes and italics:

The above response from ChatGPT achieved multiple objectives. Initially, it summarized the measurement basis in the current conceptual framework at a high level. It also indicated that the exposure draft added the role of entry prices and exit prices in financial reporting; this has not been incorporated in the conceptual framework before, so it is significant that ChatGPT accurately noted this change. Finally, ChatGPT accurately noted that the exposure draft aligns well with the conceptual framework’s fundamental characteristics of relevance and faithful representation.

The authors further asked ChatGPT which industries would be most impacted by the proposed measurement change. ChatGPT responded with the following sectors: real estate and construction, financial services, energy and utilities, and technology and pharmaceuticals. ChatGPT also listed its underlying reasons why these four industries will be most affected. For example, it provided the following reason for the real estate and construction industry:

“The shift towards a more explicit use of entry and exit prices in the exposure draft could impact how property, plant, and equipment are valued on the balance sheet. Real estate companies often deal with assets that may be revalued based on current market conditions, making the measurement changes particularly relevant.”

Next, the authors asked ChatGPT to act as a company’s accountant and draft a memo to the CFO of a real estate and construction company, informing them about the potential impact of the proposed changes. ChatGPT detailed the impact in the following four ways: property valuation changes, impact on debt and liabilities, changes in depreciation and amortization, and transparency and comparability. It also gave its reasoning why and how the change of measurement will impact financial statements in each specific area. For example, in the area of property valuation, ChatGPT stated:

“The draft suggests a shift towards more frequent use of current market prices (exit prices) for asset valuation. Given our substantial real estate holdings, this change could lead to greater volatility in reported asset values, reflecting more immediate market conditions rather than stable historical costs.”

ChatGPT then went beyond the analysis of the four affected areas and suggested that a CFO should prepare for the change in three directed channels: financial systems review, stake-holder communication, and engagement with auditors, and offered a brief explanation of how to accomplish them. Finally, ChatGPT ended the memo with the following conclusion:

The authors proceeded to request from ChatGPT detailed instructions on how to prepare for the change. In response, it provided comprehensive steps for the three channels in the CFO memo and introduced three additional items: training and development, policy and procedure updates, and monitoring and evaluation. It concluded with the following reasoning:

“By following these steps, you can ensure a systematic approach to preparing for the changes proposed in the FASB’s exposure draft, minimizing disruptions to financial reporting and maintaining compliance with evolving standards.”

Overall, this example demonstrated that ChatGPT 4.0 can read an exposure draft and accurately spot the changes to existing standards. It also provided what the authors deemed adequate reasoning behind its suggestions on how to prepare for the change. Therefore, in this example, the authors believe that ChatGPT 4.0 can reduce accountants’ workloads and help them provide reasonable and sufficient guidance on changes to GAAP.

In this example, the authors believe that ChatGPT 4.0 can reduce accountants’ workloads and help them provide reasonable and sufficient guidance on changes to GAAP.

Evaluating Going Concern Status

Auditing standards require that auditors assess whether there is substantial doubt about an entity’s ability to continue as a going concern for a reasonable period (AS 2415, PCAOB, 2015; formerly SAS 59). This period should not exceed 12 months beyond the date of the audited financial statements. Evaluations of going concern are based on qualitative and quantitative information. This information pertains to relative conditions and events that are known or reasonably knowable at the time the evaluation is made.

The authors asked ChatGPT to provide guidance on the going concern status of Apellis Pharmaceuticals for the fiscal year ended December 31, 2023. We uploaded the unaudited 2023 third quarter 10-Q to ChatGPT 4.0. We also uploaded the 10-Ks for 2022 and 2021. The authors intentionally avoided using the 2023 audited 10-K, as it already contained the auditor’s opinion on going concern. ChatGPT 4.0 is trained based on data until March 14, 2023. By this date, Apellis’s 2023 10-K had not been issued, which avoids any look-ahead bias.

Similar to the prior example, the authors assigned ChatGPT the role of an external auditor. The prompt was as follows (full transcript can be found at https://tinyurl.com/3ecnnd4m):

You are assigned as a new auditor of Apellis Pharmaceuticals, Inc. You are responsible for auditing the company’s financial statements for the year ended on December 31st, 2023. Prior auditors of Apellis had issued going concern opinions in the auditor’s report in the last couple of years. Please read Apellis Pharmaceuticals 2023 Q3 quarterly filings uploaded here and give me a summary of the going concern risks that Apellis is facing.

ChatGPT took Apellis Pharmaceuticals’ 2023 third-quarter10-Q and highlighted the following points: First, the consistent net losses and significant cash burn are red flags for potential liquidity risk. It noted that, according to the MD&A section, Apellis’s management is confident that its $452.4 million in cash and cash equivalents will sustain operations and capital expenditures through at least the second quarter of 2025. Second, it noted that, according to the going concern section of the 10-Q, the company’s viability hinges on the success of its two drugs, Syfovre and Empaveli, which will impact its ability to secure financing. Overall, ChatGPT successfully identified specific areas in the 10-Q that affect the company’s going concern status and provided an accurate analysis of associated risks.

The authors then asked ChatGPT to suggest audit procedures to evaluate the company’s going concern status for its annual audit. ChatGPT suggested ten steps: 1) understand management’s assessment, 2) analyze financial statements, 3) cash flows and forecast review, 4) examine external financing options, 5) review compliance and regulatory environment, 6) consider operational and market conditions, 7) management plans and other actions, 8) subsequent events review, 9) audit documentation, and 10) consideration for audit opinion. For each step, it gave an overview of the actions to be taken to carry out these procedures. The first two of the ten suggested procedures are noted below and were used in the following conversation with ChatGPT:

The authors requested more detailed suggestions from ChatGPT 4.0 regarding the initial two suggestions. For conciseness, we did not ask for detailed suggestions for all ten procedures. Interested readers can employ comparable prompts for the rest of the eight procedures. For the first step, “Understand Management’s Assessment,” ChatGPT was given the following prompt based on the ChatGPT output above:

Thank you for the above ten steps. For the first step, prepare a memo to the Apellis CFO asking for their assessment of the company’s ability to continue as a going concern. Their assessment should include specific details on the assumptions and criteria they use to determine the company’s ability to continue as a going concern, including cash flow forecasts, budgeting, and financial projections.

ChatGPT drafted a request memo outlining the specific documents and information that would be required from management. This included a detailed list of documents, such as assumptions and criteria used, cash flow forecasts, budgeting and financial planning data, financial projections, underlying data, sensitivity analysis, board minutes, and other documentation. The authors found this list to be comprehensive and believed it covered the documentation that would be needed for a successful audit.

Lastly, the authors asked about the second step suggested by ChatGPT:

For the second step in the audit procedures above, you need to analyze financial statements. Attached are the Q3 quarterly reports and last two years of annual reports. Specifically, you need to do two sub-steps. First, perform analytical procedures on the historical financial data to identify trends and fluctuations that might indicate potential going concern issues. Second, scrutinize the balance sheet for signs of liquidity issues, such as increasing current liabilities, declining current assets, or significant debt maturing within the next year.

ChatGPT responded with a breakdown of the net loss analysis, cash flow from operations, current liabilities vs. current assets, accounts receivable and inventory, cash and cash equivalents, and long-term liabilities. ChatGPT’s analysis doesn’t rely on a single fact, but rather on various components. While the current ratio analysis provides valuable insights, it is not the sole indicator affirming the company’s stability. This analysis warrants consideration within the framework of broader economic circumstances, suggesting a thorough investigation. Investors and stakeholders should conduct a thorough investigation to assess the company’s financial resilience and long-term viability. ChatGPT concluded as follows:

“Apellis Pharmaceuticals shows signs of potential liquidity risk due to consistent net losses and significant cash burn. Although current ratios appear healthy, the rising accounts receivable and inventory levels, alongside a decrease in cash reserves, may signal underlying issues in cash flow management and sales efficiency. This analysis should be considered in the context of broader economic conditions and sector-specific challenges, and further detailed investigation is recommended to explore these trends more deeply.”

In summary, ChatGPT can leverage its ability to sift through lengthy documents and large historical data quickly. Its examination of Apellis’s 10-Q and 10-Ks can provide auditors with a fresh viewpoint. This analytical approach can help auditors think creatively and critically. By utilizing ChatGPT’s analytical tools, auditors can enhance their work more efficiently.

Automating the analysis of the 10-Q enables accountants to redirect their focus towards more strategic endeavors, such as financial planning, analysis, and advisory services. This shift can not only enhance productivity, but also foster greater client engagement. AI can enhance client interaction by streamlining processes and delivering insights promptly and accurately. Effective client communication serves to delineate a company’s objectives, strategies, and associated business risks. This can also help identify potential risks that could lead to material misstatements in the financial statements.

The Benefits of AI

As tedious tasks are increasingly able to be performed by AI, accounting work that consists of repetitive tasks can be phased out. One factor contributing to college graduates’ reluctance to enter the accounting profession is the perception that it is dull and unappealing. But AI presents the opportunity to automate many manual tasks, freeing professionals to engage in more creative and strategic roles.

Accountants who possess the skills to effectively utilize AI may be able to outperform those who do not.

With the automation of data collection, processing, and analysis, AI can conduct many aspects of the more daunting data wrangling tasks typically handled by junior accounting staff. These include “inputting and matching data, receipt reconciliation, creating and sending invoices, expense reports, tracking price changes, account reconciliation, sorting transactions, data recording and reporting” (“Artificial Intelligence in Accounting: What Will Happen to Accounting Jobs?” CMA Exam Academy, https://tinyurl.com/ncszbdvh).

AI can improve the data processing capabilities previously handled by accountants, promising greater efficiency and lower error rates. AI can also enhance auditing and financial reporting processes. AI-powered tools might be used to keep track of vast amounts of financial transactions and send alerts of problematic patterns and abnormalities to auditors. This proactive approach reduces manual efforts, improves the probability of fraud detection, and mitigates potential losses.

Once many repetitive tasks are taken care of by machines, AI can broaden accountants’ perspective on their work by using data from a wide variety of sources. This collective intelligence could draw on data from across sectors and jurisdictions, providing creative and potentially more critical viewpoints on the questions accountants explore.

Finally, the current versions of many AI services can explain their reasoning, logic, and the data sources underlying their outputs. This allows accountants to easily check and verify the thought process, logic, and data sources behind the output.

Accountants who possess the skills to effectively utilize AI may be able to outperform those who do not. It may take a significant amount of time and training to use AI efficiently. The impact of AI on various organizations can be multifaceted, as not all organizations are similarly ready for the breadth of AI implementation. In the authors’ opinion, accountants will not be entirely replaced by AI, but rather accounting work will be transformed by AI. While tedious and repetitive tasks are increasingly accomplished by AI, more opportunities in the strategic and creative roles of accounting will be available. This shift will allow accountants to focus on higher-value activities that require critical thinking and problem-solving skills, ultimately enhancing their overall contribution to a business.

Admittedly, while AI can help auditors decipher large documents and data sets, it is not flawless. AI cannot make judgements that require human experience, ethics, and intuition. Accountants are needed in situations where decision making is complex and ambiguous. AI may need help understanding the context behind certain transactions or decisions. In such cases, AI might give an unclear or inaccurate answer or even hallucinate unreal facts, making human interpretation necessary. AI also lacks the emotional intelligence and personality needed to foster client relationships. Due to its inability to update instantaneously, AI systems require continuous maintenance to remain current; relying solely on AI can lead to erroneous conclusions. While AI offers tremendous potential benefits, addressing these drawbacks requires careful consideration, ongoing research, and responsible deployment to ensure that AI technologies are developed and used in ways that benefit society as a whole.

Ariana LoBianco is a student at St. John’s University, Jamaica, N.Y.

Kevin Jialin Sun, PhD, CFA, is an associate professor of accounting at St. John’s University.

Xin Zhao, PhD, is an assistant professor of accounting at St. John’s University.