Artificial intelligence (AI) has captured the attention of financial executives, both in large and small organizations, looking to take advatnage of the productivity gains promised by the new technology.
But there are both opportunities and risks associated with adopting AI, and in a profession that demands consistency and accuracy, there remains significant questions about the road ahead.
As part of planning Financial Executives International’s recent virtual event on AI In Finance and Accounting, the Financial Education and Research Foundation (FERF) conducted a survey of FEI Executive Members to get their perspective on the burgeoning AI phenomena.
In addition, FERF conducted two “case study” interviews with KPMG and NetSuite regarding their roadmap for AI technology into the not-so-distant future.
Also included in this report is an interview with one leader who is in the midst of implementing their own large scale AI project within the financial reporting function.
This research brief includes:
Member Survey: Setting Expectations in AI in Finance and Accounting
Respondents to the survey regarding AI in finance and accounting sent by FERF in February 2024 were cautiously optimistic regarding the future of technology, with majority having a cursory experience with large language models outside of consumer uses of ChatGPT.
That said, those financial executive who applied AI in the context of their processional expertise understood its significant potential.
When asked about initial use cases, answers included:
- “Mostly ChatGPT to help me with writing presentations or upgrading software with better AI capabilities.”
- “Assist with accounting research, policy writing and job description modifications. Minimal use in data analysis due to security.”
- “Used AI to assist in guiding new employees to our policies and recommended training. Now we are seeing this used for data mining.”
- “Only to the extent that it exists in some primary tools already. We have not deployed any new or standalone AI packages. We are making use of ChatGPT as a training and data resource.”
AI Case Study 1: KPMG
Key Takeaways
- A First Step Internally, Then the Audit: Today AI is being leveraged to enhance overall KPMG operations and internal policy management, to increase productivity and free up resources with an eye on the audit processes.
- Generative AI as a Solution: Generative AI is poised to alleviate the burden of mundane, repetitive tasks in accounting and finance and, by extension, within the audit process. This will allow external auditors, internal auditors, and other skilled finance professionals to focus on more “value-added” activities.
- AI Will Redefine Offshoring: The traditional model of offshoring audit work could be redefined as AI capabilities increase, potentially replacing offshore centers built over the past two decades.
Rationale for AI
Advancements in artificial intelligence (AI) are forging a new era in the audit and accounting industry, reshaping the traditional roles and redefining the required skills, says Edward Moran, Managing Director, Audit Tech – Innovation, KPMG US.
As part of KPMG’s initial AI implementations at the firm, Moran says he is finding that automating routine and time-consuming tasks with AI can free up professionals so they can focus on more complex and value-added audit work. This will be particularly critical for those new professionals entering the workforce with Certified Public Accountant (CPA) credentials who are searching for more strategic responsibilities.
The result is that staff in public accounting and within industry should prepare for a workforce paradigm shift that will emphasize strategy over tactics and insight over repetition. “Their skills are better spent elsewhere than checking two documents, comparing two numbers, and summarizing documents,” he says, adding that professional skepticism remains essential to the auditor’s craft.
But there is a burgeoning need for data literacy fueled not only by AI but by the plethora of data analytics tools that are integral to modern enterprise resource planning (ERP) systems.
Moran acknowledges the current talent shortage in accounting, emphasizing that the allure of an accounting career has diminished over the past several years and pointing out the struggles audit firms face in recruiting individuals equipped with the necessary skills.
However, he is optimistic about the potential of generative AI to alleviate some of these issues by shifting the career track. When discussing the difficulty of acquiring data talent to build AI processes within public accounting firms, Moran describes it as “challenging,” as everyone is vying for individuals with these capabilities, citing a program called the Masters of Accounting with Data and Analytics (MADA), which has been co-created with accounting programs across the United States to tailor the curriculum for the evolving needs of the industry. Moran highlights the firm’s strategy of selecting individuals for the MADA program and subsidizing their degrees as they commence their careers with the firm.
Confronting current accounting programs in colleges to address the intersection of accounting and technology, he calls for a significant shift. He recalls past dialogues with academics who saw computer science and data as distinct from accounting education. “That’s not in the accounting program. That’s a different school,” he reminisces on the initial resistance to integrating these disciplines. However, he notes a positive change, with universities now fostering multidisciplinary programs that blend computer science, data analytics, and accounting, supplemented by communication classes aimed at enhancing storytelling skills.
Finally, Moran touches on the future of offshore centers in the context of AI’s rise. He hasn’t witnessed a takeover yet, but he recognizes a “strong logical argument” for AI to assume tasks traditionally outsourced offshore. Generative AI, with its capability to compare, summarize, and synthesize research, could potentially render offshore practices obsolete. “It can take place in software instead of in a person’s head,” Moran concludes, contemplating the broader implications for the offshoring industry.
Moran’s perspective captures a transformative period in accounting where AI is not just an enabler but a catalyst for a complete overhaul of traditional practices, demanding new skill sets and reshaping career pathways. The industry is on the cusp of embracing AI to enhance efficiency, focus human talent on critical thinking and analysis, and redefine the educational curricula to prepare the next generation of accountants.
Solutions
KPMG’s journey into AI began with its leveraging OpenAI’s technologies to enhance various operational aspects, notably including the writing of SQL queries. Moran shares, “I could take an auditor who knew very little about writing SQL queries and very quickly get [them] up to speed by using generative AI.” This initial foray into AI was not just about operational efficiency, however; it also highlighted the potential for AI to facilitate complex tasks such as contract review, extracting key terms from leases, agreements, and more.
Beyond the immediate applications in auditing, Moran envisions a broader role for AI within the firm’s operations. He points out how AI can revolutionize knowledge management and internal policy adherence, saying, “You can put those things in generative AI, and you can have everyone have access to the most updated policy and be able to ask a question.” This functionality extends beyond mere operational efficiency; it encompasses a fundamental shift in how information is accessed and utilized within the firm, enhancing both employee engagement and productivity.
Moran’s vision for AI in audit extends to improving the quality of audits themselves, where AI acts as a copilot to auditors, assisting in the identifying critical data points and anomalies within vast data sets. “We’re using it right now to look for anomalies in large data sets,” Moran states, emphasizing the shift from traditional sample-based audits to more comprehensive data reviews facilitated by AI. This methodological shift promises not only enhanced accuracy but also a more profound confidence in audit outcomes.
Discussing the potential for AI to compare a client’s financial disclosures against broad industry standards, Moran illustrates AI’s analytical prowess: “I can take my client’s 10K and say, ‘Hey, look at all of FASB’s guidance and tell me if my client’s 10K… Does it have all the disclosures in it [that] it has to have?’” This application of AI in auditing represents a significant advance in ensuring compliance and best practices, leveraging AI’s capability to digest and analyze vast quantities of data and regulatory material.
Moran is acutely aware of the cultural and operational shifts required to integrate AI fully into auditing processes. He underscores the importance of change management and the need for staff to embrace AI as a tool that enhances rather than threatens their roles.
Looking ahead, he acknowledges the challenges and uncertainties of fully integrating AI into auditing, particularly around auditing internal controls and regulatory compliance. However, he is optimistic about the role AI will play not just in transforming audit processes but also in shaping corporate governance and control environments. “They’re going to need controls around their AI… Did they use the AI to write the control around?” Moran muses, highlighting the recursive nature of AI’s impact on business processes and the audit profession.
As KPMG navigates the early stages of this transformation, Moran’s narrative underscores the potential and the challenges inherent in integrating AI into auditing, pointing to a future where AI is an indispensable part of the audit process, driving improvements in quality, efficiency, and reliability.
Impact, Learning, and Advice
Moran says the dynamics of how AI is reshaping the audit process means that both auditors and preparers need to rethink their approach to the entire universe of available financial data and how it can be leveraged. That translates into the importance of clean data for leveraging AI effectively, and the critical role of human oversight in mitigating AI-related biases and errors.
He begins by addressing the shifting landscape from the preparer’s perspective as AI technologies advance, drilling deeper into financial data and its immediacy.
“The data is there and it’s there quickly. A transaction happens; journal entry is booked; it’s there,” he explains.
This immediate data availability sets a new standard for auditors, urging a move toward a more proactive and frequent auditing approach. Moran envisions a future where the traditional, scheduled audits give way to a dynamic, ongoing scrutiny facilitated by AI.
Discussing the impact of AI on continuous auditing, Moran is both realistic and optimistic. He acknowledges the hype around AI but emphasizes its tangible benefits: “AI is a great data and analytics machine. Is it perfect? No.” However, he clarifies that the revolution in auditing is less about AI itself and more about the availability and quality of data, remarking, “I think that continuous audit would be helped by AI. But I think more [important] is the availability of clean data that modern enterprises now have.”
On the topic of standardizing data exchange between preparers and auditors, Moran sees significant opportunities in leveraging ERP systems to enhance their role. This proactive approach could revolutionize how companies monitor their financial health and how auditors engage with their data, potentially making audits more efficient and insightful.
Moran is is both realistic and optomistic when discussing AI's impact on audting.
“AI is not a truth machine,” emphasizing the importance of human intervention in the AI-driven audit process. Moran articulates the nuanced reality of relying on AI: “It can be wildly, wildly wrong, and it is a lot. But that said, it’s still one of the most amazing tools that we have available to us.” This highlights the delicate balance between harnessing AI’s capabilities and acknowledging its limitations.
As for advice to clients on navigating the future of AI in auditing, Moran stresses the importance of beginning with low-risk applications and gradually building an organization’s AI competence. Moran encourages engaging external expertise to navigate the nascent field.
Enterprise AI Q&A
FERF spoke with the leader of one organization to get their thoughts on implementing artificial intelligence across the enterprise and in finance and accounting.
FERF: Can you outline the specific areas within the finance process where artificial intelligence is having the biggest impact withing you organization today?
Financial Executive: Our company is currently exploring the usage of multiple Generative AI use cases within finance.
Primarily we’re focused on using Gen-AI to generate automated insights for our financial reporting and using chatbots to assist in querying our structured data, both within horizontal applications across finance. We are also using AI to identify potential discrepancies in our data to reconcile data across our ledger systems.
FERF: Could you describe where AI could have the most significant impact in terms of efficiency gains and risk reduction in finance in the next five years?
FE: AI has the capability to virtually eliminate errors in our reporting, working hand in hand with our finance professional, these tools are likely to transform and streamline the previously painful manual process that go into creating our reporting.
FERF: What are the key performance indicators (KPIs) that you would use to evaluate the success and effectiveness of AI implementation?
FE: We first evaluate AI in terms of return on investment on our strategic objectives.
For instance, we’ll be tracking time saved by discrepancies in our ledger data being identified and remediated, having the human-in-the-loop only doing review. To this purpose I would imagine mean time to repair (MTTR) will be a key indicator, but as we increase our inventory of use cases, I expect other KPIs and success metrics will come online.
FERF: Are there specific areas of accounting and financial reporting where AI is currently being implemented within your company?
FE: Non generative AI -- i.e. machine learning models -- have long been a part of our forecasting, testing and close procedures. We’re currently evaluating how Generative AI can be effectively and safely integrated into our processes.
FERF: What strategies would you propose to mitigate potential risks associated with the use of AI in auditing, such as data privacy breaches or algorithmic biases?
FE: Our firm takes the risks posed by AI very seriously and as such have established a robust governance and review process for any and all AI implementations.
Models are monitored on an ongoing basis, tested regularly against our tolerance criteria. Our business units are partnered with our technology partners, data teams and enterprise risk partners to ensure use cases are fully vetted.
AI demands an atypical strategic level of cooperation and collaboration across all your business partners, risk partners, and technology.
You cannot develop AI in a silo. You must be unified against the problem. Furthermore, it’s important that you’re constantly reviewing your audit procedures to adapt to new risks as they appear.
FERF: What types of data analytics tools and techniques would you suggest integrating with AI to enhance the effectiveness of audit procedures and uncover insights from large datasets?
Most tools in the market are reacting to the AI moment, and many are integrating AI capabilities into their existing toolkit. We’re currently evaluating the usage of several AI tools relating to machine learning.
However, much of our development modeling is still done in traditional programming languages such as R and Python.
I expect that will shift as AI enabled tools and intelligent automation become more widely accepted.
Case Study 2: Oracle NetSuite
Key Takeaways
- Holistic AI Integration: NetSuite is committed to embedding and evolving AI across its ERP suite.
- Current AI Applications and Productivity Patience: NetSuite is employing generative AI and natural language processing models across all of its ERP applications, including its core financial management and accounting services. Users can see productivity gains in the context of real-world workflows and get better insights over time as the models continue to learn on the data.
- Data Governance: NetSuite continues to prioritize data privacy and security, giving customers tight control over data management and security.
- Challenges and Future Directions: With the advent of data and analytics tools, there's a growing need for data literacy among accountants, who still maintain a healthy professional skepticism of AI technology.
Rationale for AI
NetSuite has been incorporating AI capabilities into its features for years and continues to take a “holistic” approach to solving customers’ issues with the technology, says Lisa Schwarz, Senior Director of Global Product Marketing for the Oracle NetSuite Global Business Unit.
“With NetSuite there is no separate implementation of AI, it is — and will continue to be — embedded in features to assist and advise users. If it is a new feature that has AI capabilities, then it’s just part of the overall feature functionality.”
Given NetSuite’s iterative approach to the technology, the ERP provider plans to continue to explore AI technologies to bring its customers more AI-driven business insights and automation as processes and use cases are presented.
“We want customers to adopt AI at their own pace because a lot of companies are still trying to figure out policies for AI use in their own enterprise.”
The company remains practical about the future of AI and its role in ERP, especially as finance and accounting users move up the learning curve.
“The response to AI thus far from customers has been very positive overall, especially in the initial implementations of NetSuite Bill Capture, which populates the data of the scanned invoices into the NetSuite bill record,” Schwarz says. “The more use of the technology, the better it performs because AI is continuously learning.”
Solutions
Schwarz laid out the specific areas of NetSuite’s applications that currently employ AI technologies — such as machine learning and natural language processing — and the key business functions they impact.
Advisor capabilities for informing decisions:
- Supply Chain: The NetSuite Supply Chain Control Tower uses AI to run simulations that track inventory levels and predict the effects of changes to, say, a bill of materials or production process. Predictive risk tools let planners model supply and demand in detail, which can result in reduced production bottlenecks.
- Financial Planning: NetSuite Planning and Budgeting uses AI predictive algorithms to continuously monitor and analyze plans, forecasts, and variances. The solution highlights trends, anomalies, biases, and correlations so that finance departments can assess and take faster action on those insights.
- Business Intelligence: NetSuite Analytics Warehouse’s built-in AI-powered analytical capabilities help customers unveil patterns and opportunities from complex data relationships, explain key findings with chart narratives, and identify key drivers of revenue growth with predictive analytics using machine learning models. With automated analysis of operations and performance, customers can also reduce time-intensive data analysis and more quickly address business challenges.
Assistant capabilities for handling repetitive tasks
- Invoice Processing: NetSuite Bill Capture uses AI-based document object detection and intelligent data classification to bring invoice data into NetSuite faster and with reduced risk of human errors. Data is extracted from invoices and entered into NetSuite. Bill record fields are automatically populated with the appropriate details, while intelligent data recognition and information gained from previous vendor invoices help improve accuracy using natural language processing.
- Content Generation: NetSuite Text Enhance is employed in over 200 text fields that customers use to create and refine written content across the suite employing generative artificial intelligence.
- Finance and accounting: Can help expedite collections, close the books faster, and focus on more strategic and fulfilling work by accelerating time-consuming writing tasks. In addition to summarizing narratives for financial reports and personalized collection letters, new use cases include assisted authoring for journal entries to describe transactions and descriptions when creating a new account in Chart of Accounts; assisted authoring for purchase order entries including packing list and product labeling instructions; and assisted authoring for cash refund explanations to maintain consistency for customer communications and help internal auditors easily find the information they need.
- Supply Chain and Operations: Can help supply chain and operations teams streamline purchasing and logistics and improve the quality of product-related communications. In addition to suggested item descriptions, and assisted authoring of vendor engagement letters and procurement orders, new use cases help customers create support tickets for warehouse management issues and shipment summaries that describe the movement of goods; create supply chain snapshots and update supply chain snapshot simulations; and create project tasks and task assignments to accelerate project data entry, tracking, and reporting.
- Sales and Marketing: Can help marketing and sales teams accelerate tasks and create more effective campaigns that drive revenue. In addition to assisted authoring of email content for marketing campaigns and sales pitches, new use cases help customers create quotes and summarize sales events to enable more consistent engagements; assisted authoring for lead-generation communications to improve consistency in tone and structure; and assisted authoring for message development to help create targeted messaging.
- Manufacturing: Can help streamline and accelerate the management of manufacturing operations. Use cases include assisted authoring of critical data entry and tracking activities including manufacturing planned time, operations tasks, and manufacturing routing; and creating consistent definitions for manufacturing cost templates to help reduce confusion and errors when assigning them to the manufacturing operation.
- Human Resources: Can help employees, managers, and HR leaders increase the speed and accuracy of important HR activities. Use cases include assisted authoring of job descriptions and requisitions, employee goals, peer-to-peer kudos, and summaries of employee performance based on feedback gathered from peers, managers, and progress against goals throughout the year.
- Customer Support: Can help customer support agents increase productivity and improve the customer experience. In addition to assisted authoring for online comment responses, new use cases help businesses maintain accurate customer cases and issue records by summarizing customer events, root cause, and resolution.
In terms of backend processes, Schwarz states that the Oracle Cloud Infrastructure (OCI) that the NetSuite service is hosted in"leverages OCI Supercluster, which includes bare metal compute instances, ultra-low latency RDMA networking, and high-performance storage." This helps accelerate LLM training with the highest performance at the lowest cost, she adds.
Impact, learning, and advice
Schwarz also addressed concerns about data privacy, which she understands is significant for customers. "There is some concern over data, with customers particularly wanting to ensure their data is protected," she stated. This concern is mitigated by allowing customers to "adopt AI at their own pace," enabling them to become comfortable with the new technology and its implications for data privacy.
To address this concern, Schwarz pointed out several measures that NetSuite has implemented. "We have over 25 years of experience in managing and securing user data in the cloud," she explained. Each customer's AI model can be uniquely trained on their own data, and we give customers strong control over data governance and security.
Furthermore, Schwarz detailed additional security within NetSuite: "To further protect sensitive information, role-based security is embedded directly into our workflows."
This is designed to ensure that the AI only recommends content that end users are entitled to view, which enhances both usability and security. She also noted that NetSuite does not export data to other AI tools, which "would increase the risk of the data being compromised."
Member Survey: Understanding Future Risks and Opportunities
A majority of financial executives responding to the FERF survey said one of their biggest concerns around the risk of AI in finance and accounting was trust in the technology, followed by security concerns of financial data.
When asked about their obstacles to embracing AI, their focus was practical, including:
- “Client restrictions.”
- Supporting infrastructure to ensure complete and accurate outcomes.”
- “Lack of talent to take advantage of functionality.”
- “Restructuring costs.”