In a KPMG study in spring 2023, the Big 4 firm found that 77% of its business client respondents expected generative AI to have the largest impact on their businesses of all emerging technologies. However, 92% of them believe that implementing AI will introduce moderate to high-risk concerns.
For a profession as risk averse as accounting, what can we gather from this? To start, Blake Oliver, who recently presented on AI and the Profession during the Expensicon conference in Italy, noted that a paper produced by ChatGPT parent company Open AI suggests that the accounting profession is one of the most exposed to generative AI.
“This means that the time it takes to do 100% of our tasks could be reduced by 50% or more,” via use of generative AI, he said. “That’s a pretty bold statement of productivity savings.” Oliver also noted, however, that there are many areas where ChatGPT is not that adept, particularly focused on human interaction and areas of subjectivity. Whereas it excels at data-driven and analytical functions.
Oliver then asked how many of the panelists have used ChatGPT, and how often. All of the participants noted they have tried it, and 25% claimed to use it on a daily basis. The primary tasks they or their clients use generative AI for generally include:
- social media posts
- report summaries
- JASON exports
- financial analysis
- business planning
Per an article in The Wall Street Journal, further examples of the current frequent use of generative AI for business financial purposes also include:
- questions on earnings calls
- transcript analysis
- internal guidelines
- transfer pricing
- budget and actual variances
- first draft of journal entries
- reports, documents and presentations
- footnotes to financials
- auditing GL transactions
Avalara surveyed the influencers on their opinion and usage of generative AI, and 53% said they are currently using the technology for generating generic text for marketing, internal communications, policies and other functions. Other top functions they are using AI for include data analysis, scheduling and art. About a third said they are not currently using the technology but are exploring it.
Summarizing the survey data he’s seen regarding the current and pending state of generative AI adoption, Oliver noted that 50% of executives stated they are currently using AI in their finance and tax departments already, and 29% say they have plans to do so in the next year. “That means about 90% are already or expect to be using AI for finance and tax in the next 6-12 months, and the remaining respondents expect to use it in the near future.”
“There are, of course, concerns over accuracy, and hopefully ways to mitigate that, and we’ll discuss those today in the panel,” said Oliver. “But I think AI is very exciting and this is a potential solution to the accounting talent crisis caused by declining numbers of graduates, and the increased need for more accountants.”
“AI has potential to save accounting. The reason we work long hours is because computers haven’t delivered the productivity advances we were promised. We’ve just taken a lot of these paper-based processes and digitized them, but still have to do the same work.”
“Generative AI has the potential to give us the productivity boost we’ve been looking for, especially in audit and tax.” ~ Blake Oliver, CPA.
Hood then led the panelists into discussion, first clarifying the various types of AI, including generative language models like ChatGPT, as well as other models focused more on math, and forms of AI geared more toward audio or visual output.
In the survey of round table participants, nearly half of them believe generative AI tools will be used on a daily basis by most people in firms within the next 1-2 years. About a quarter of participants think that will happen within a year, while others forecast it will take at least three years.
Hood then suggested a “lightning round” discussion on, “What things do you think we could use AI for? Not limited to accounting, but in any field,” and noted some tasks that he sees as pretty obvious, such as engagement letters, social media posts, and running schedules.
“I recently saw an ad for an app that will analyze your email and your calendar to optimize the things that you have to do, based on prior behaviors. I think that will be a huge productivity boost,” said Nayo Carter-Gray, an EA and founder of the Baltimore firm 1st Step Accounting. “It will also be a big help for work-life balance and free up some brain power.”
Hood added that such an activity analysis could be useful in other areas, such as determining the areas that an individual needs to focus more on, in terms of work processes or personal life goals.
Tamara Loerzel, a partner at ConvergenceCoaching, noted that her advisory firm is implementing the work management solution Monday.com, and says they’ve been impressed with the AI built into that system, and efficiencies gained, such as being able to pull data from disparate business systems to improve project management, scheduling and coordination. “Plus, some of those routine tasks, the system is suggesting to us how to automate them,” she said.
“One of the ways I’ve found that I like to use ChatGPT is to research something I don’t know,” said Shayna Chapman, CPA, founder of the West Virginia accounting firm ShaynaCo LLC. “I’m mostly tax, bookkeeping and accounting, and don’t do much A&A work. So, I don’t need specific FASB footnotes and things that go into audit and financial statements, or how they should be specifically presented. ChatGPT helps me see and understand how these things should be done. It also helps if you have a short attention span on some topics, it helps ADHD people get more organized and get started with managing tasks and getting organized.
“I’ve worked in large and small firms, and they all struggle with breaking down technical financial information into layman’s language for their clients,” said Gail Perry, a CPA and editor-in-chief of CPA Practice Advisor. “I would think that this would be a good place to use ChatGPT or another generative AI.”
Diane Yetter, founder and president of the Sales Tax Institute, is using the technology for course development, stating that it is much faster and good at summarizing information. However, she noted that she’s leery of allowing the AI systems to have access to proprietary information like her emails.
Caleb Jenkins, an EA with RJL Financial, said he loves generative AI for “anything of a communicative nature,” and shared how he’s using it for several routine tasks, including creating custom out-of-office messaging.
“AI is very important and an area of continued investment for Avalara. We’ve already used AI to make transformative headway in areas like document management and tax research,” said Sona Akmakjian, CPA, global head of strategic accounting partnerships for Avalara. Not that long ago, she says that tax technology companies had large staffs of attorneys and CPAs who each had a jurisdiction to research. “With the local levels, it required personal relationships within those jurisdictions because websites weren’t kept updated. So it was a very manual process to keep tax databases up to date. We now use AI to streamline the tax research maintenance, and there is less labor required and improved accuracy.”
On the academic front, Annette Nellen, CPA, CGMA, Esq., a professor in and director of San Jose State University‘s graduate tax program, noted that she cautions firms to fully explain their policies regarding AI when it comes to new hires, particularly when it comes to confidentiality and other issues.
“In talking to my graduate students about it, I was surprised by what they thought were the good and bad points, and that some hadn’t explored it at all.” She suggests they consider using AI as similar to speaking with a colleague. Ask it a question, and don’t take it as the final word on the matter, but as an additional point of information, an additional resource. “But you are still responsible for figuring out if the information is correct or not.”
“The way I look at AI is that it’s kind of like DOS, a command line. But we’re really at the beginning of a revolution. The start of another wave of change, and that’s why there’s a little hesitation and fear,” said Peter Wen, CPA, the founder and CEO of Tallyfor. “The challenge will be ‘what is the interface going to be?’ Right now you have to learn how to use it, but asking it questions can still be difficult. Our job on the technology side is to make the interface more user-friendly, to make it more like the Windows of ChatGPT. So that’s what Tallyfor has been working on.”
W. Michael Hsu, CPA, founder of CFO consultancy DeepSky added his thoughts on the quality of the information obtainable from generative AI systems.
“I’ve spent 400-500 hours on ChatGPT since it came out last year and it does tremendous things, but the issue is how do you get the knowledge out,” he said. “And once you get the answers, you have to have the knowledge to identify when it gets some things wrong. You have to be knowledgeable on the topic to get the most out of it.”
“Generative AI is only as good as the user themselves.” ~ W. Michael Hsu, CPA.
The group collectively agreed that AI will not replace accountants, although it may replace some tasks, and as Hsu noted, those professionals who embrace it earlier and become more adept at its use will elevate themselves far above the abilities of those who do not become as proficient.
What Else Can AI Do?
Dan Hood then led the group towards discussion of other areas that generative and other forms of AI may be adaptable to, or may already being implemented in.
“In addition to text-based generative AIs like ChatGPT, there are other models specifically tuned to math, imagery and other formats,” he noted. “So, looking beyond ChatGPT, are there other applications that are not prompt-driven, meaning you ask it a question or give it a command? Are there any that you can just point at a collection of data, images, or other content, and tell it to absorb it and report?”
Nayo Carter-Gray noted the use of AI in scanning medical imaging to find correlations and trends, which can be useful in diagnoses and treatments.
Blake Oliver referenced chat-based customer service companies, many of whom have developed their own support databases and systems that list pre-canned responses based on customer need, with agents able to select the one appropriate to the situation.
“This would be a natural use of ChatGPT, where it could automatically send one of those responses based on customer questions or comments.” He suggested this could also apply to accounting firm client service, too. For instance, when a client emails asking for a tax return or financial statement. AI could find that information and send it to them via secure portal.”
Diane Yetter sees potential for the use of AI in sales tax processes. “From a sales tax perspective, we’ve heard of more bot-based AI pulling invoices to support an audit. Some challenges on filing returns include how often states change their websites and portals. Can we now get the AI to read the invoices and analyze if the transactions are what they say they are, and whether they are taxable? This is a time-consuming area of sales tax audits that I’d love to see AI do.”
Caleb Jenkins noted a company named Digits, which has been doing the predictive side of analytics using AI tools and has combined that with generative AI as well. This lets them connect to GL sets to automatically recommend and suggest cleaning up transactions.
Randy Johnston, the founder of Network Management Group, Inc. and executive VP of K2 Enterprises, noted some interesting applications of AI have included the use by a farm support group that has implemented chatbots and reduced human support contact by half. “The amount of cost-savings on chat support agents alone is pretty stunning.” He further noted that “building AI into technology platforms is a big deal.”
“We’ve seen AI around for a very long time. The bigger part we’re going to see as we get better and tighter integration is actually being able to control the content and the data that’s being researched,” said Dan Luthi, CPA, a partner at Ignite Spot Accounting. “I think that’s one of the things that will make a massive change in the evolution of AI for us as accounting professionals and also for consumers.”
“The biggest struggle right now is that it’s using a huge database that it’s evaluating from, which creates a difficult context for value-driven functions,” he added. “It’s the reason why AI art is so skewed, it’s looking at Monet, and DaVinci, and others, and fingers are different in all of them. It’s the same thing when it comes to creating communication for a client or evaluating financials.”
Luthi referenced a session at the 2023 AICPA Engage conference that discussed the topic of setting the database for generative AI to research, instead of having it use the generic data bank that ChatGPT and current consumer models use, which is essentially all website information prior to 2021.
What are the biggest opportunities for AI in accounting?
Dan Hood then directed participants to consider what functions they see professionals benefiting from with AI in the current, near or more distant future, whether they are “pie in the sky” dreams or more realizable.
“AI is going to be great for training,” said Nayo Carter-Gray. “These how-to AI’s can hold large data sets that can teach, maintain formulas, or help staff remember how to do certain tasks. It will free up our brain power, freeing us from a lot of tedious work, and spending that time with clients and on more profitable functions. If we can have the AI doing forecasts and other projections, we can use that to further our client relationships.”
Hood agreed with these points, noting that AI-based training will also further enable more organizations to take advantage of remote workforces, since training won’t be exclusively an in-office function. These training databases could be built in-house by firms, or via adoption of best practices that could be developed by professional societies and vendors.
When asked what the first impactful uses of generative AI in accounting firms will be, attendees were basically split into three groups, with equal numbers suggesting automating data analysis, marketing functions, and handling client communications.
Caleb Jenkins sees a near-future benefit coming when multiple apps are integrated with AI, so that it can pull data from whatever systems are necessary, and then auto-respond to client requests, or help generate forecasts, financials, and other communications. “This may be a few years, but I’m currently using it for proposal follow-ups, with details of all of their services stored in ChatGPT, so that these routine client communications can be handled more easily.”
“I think AI will give us a new service line. People are willing to pay for our expertise, and we will be able to facilitate more projects for clients, as well as opportunities that we may be hesitant to take on now,” said Nicole Davis, founder and principal at Butler–Davis Tax & Accounting, LLC. “I think it will help us better provide advisory services, in ways that can be scaled to any size firm.”
Dan Luthi noted that he sees some things that should be very simple for AI to do, even with today’s technology, but would be very useful for accountants if they could be automated.
“How amazing would it be to have your bank feeds re-evaluating against data out on the market, to see how others have coded similar transactions from the same or similar vendors, and having transactions automatically classified,” he said, adding that the data would then be confirmed with the client. “That’s a super basic, simple function that would be really valuable in cutting down the back and forth between the firm and clients.”
From a business development and growth perspective, Tamara Loerzel sees firm partners being able to use AI to better analyze their own clients. “I know some firms are doing this to look at cross selling opportunities, proposing additional services, and even generating communications touch points to better serve clients and customize their experience.”
“This may be a pie-in-the-sky hope, but how great would it be to never hear another word about Excel? Ever. No articles. No clients using it, it’s gone. That would free up so much time for accountants all over the world,” said Amy Vetter, CPA, CGMA, CEO of The B3 Method Institute and Drishtiq Yoga.
Mark Koziel, CPA, president and CEO of Allinial Global, was the last to comment on the coming applications of AI for firms, noting that Dan Luthi’s hoped-for AI automation and integration of bank feeds is already starting. “There is a company called Audit Sight that is doing a lot with bank feeds to automate a lot of the audit steps around receivables and inventory, and we already have several of our firms using it. The next step is to bring this technology into the client accounting space.”
Data Security, Copyright and Other Concerns
Among professionals, security and ownership of data uploaded into ChatGPT is a highly watched topic, as there are questions as to who then legally owns or has the right to use that data.
“As a company that produces content, that we copyright, I wonder how you control your copyright if someone else puts your copyrighted information into a generative chat platform?” said Diane Yetter. “Is copyright protection a thing of the past?”
Nayo Carter-Gray said that she stresses data security with her clients, and this is another area where she will need to advise clients to be safe. “My biggest concern is people using ChatGPT sites to produce works and then selling them, even though the items were created with AI-generated data or images, that were basically sampled by other people. Additionally, will the generative AI companies come forward in the future and claim ownership of the works produced with their systems?” She also voiced concern over whether use of AI will have an effect on firm insurance as it regards data security.
Dan Hood agreed with the data ownership concerns posited by Carter-Gray, and noted that firms need to start developing rules for how their staff use the systems, from data input, through use of the output information. Further, he suggested users treat the output as additional research from which they can reference or use to create their own unique content, much as they might with a reference book like an encyclopedia.
Kelly Phillips Erb, a tax and finance writer for Forbes, noted that Sarah Silverman is suing ChatGPT and Meta for using her book in their training manuals. She said that she had never consented to this use, which has now been disseminated via the systems. Blake Oliver noted in the roundtable’s accompanying chat discussion that the U.S. Copyright Office has ruled generative AI output is not protected.
Shayna Chapman noted she is concerned about the vulnerability to cyber-attacks, and the location/country where data is being stored, and whether that information would have to be disclosed to clients. “At this point, I’m not putting any client identifying information into these systems. I just don’t know what I don’t know yet.”
“We’ve used AI in a lot of situations to create efficiencies and to automate processes,” said Jenni M. Huotari, CPA, partner and business outsourcing practice leader at Eide Bailly. “One of the obstacles we’ve found is people just going through the motions of doing processes, and not understanding the why. There’s been a disconnect between the learning aspect and really understanding. So how do we get our teams to really understand, and get AI to help us increase the learning and understanding of processes?”
“ChatGPT is powerful and it’s probably here to stay. It is critical for firms to develop guidelines on the use. What concerns me is the risk of staff using it as a source of truth without validation and internal policies or guidelines,” said Sona Akmakjian. “The accounting profession also has to address privacy, confidentiality, and integrity. That’s why Avalara is implementing policies to give employees guidelines on how and when to use the technology . Firms need to develop similar guidelines.”
To What Extent Do Accounting Firm Leaders Need to be Concerned About AI?
Dan Hood then moved the discussion toward the day-to-day issues that firm leaders will face with AI, noting that adoption by the profession may be similar to its adoption of the cloud. Some firms moving into it more intentionally, while others took a gradual and hesitant approach as the technologies that they used started to include cloud components or evolved into web-based systems. In the end, almost all firms are now mostly in the cloud, whether or not they took intentional action to go there.
With AI, Hood questioned how active firms will need to be in adopting the systems, or whether the functions will end up being integrated into the existing tax, accounting and business applications they already use. “How much will firms need to proactively adopt, and when, and over what time frame?”
Dan Luthi suggested that the pace of change in the technologies, and the scope of capabilities and adoption will change dramatically even in the short term.
“The biggest thing to me is there needs to be better education on creating the prompts [that staff use to direct the AI systems], and firm owners have a responsibility to help staff learn how to ask the right questions” he said. “These don’t need to be directly related to accounting and financial information, but just getting staff familiar with how the systems work, and how they communicate. That will help us better prepare for the wave of the really deep value-added AI tools that are going to be coming out, because the team will know how to ask questions. It’s no different than teaching staff how to research or use other office technologies.”
Diane Yetter stressed that firm leaders need to set the policies and educate staff on what the firm sees as appropriate use of the technologies. “Leaders need to be aware that some of their staff, particularly younger members, may already be exploring these tools on their own, so there needs to be understood guidelines. Being aware of how different people are using it and monitoring the usage.”
Hood recapped the major action steps the panelists had identified regarding AI that firms should take now or very soon:
- Find out if AI is currently being used at all in the firm.
- Set policies for its usage.
- Develop the skills to interact effectively with these technologies.
I Really Wish AI Could Do …
Hood had one final question to the panelists, asking, “Is there a single use of AI that you think would be most compelling to make your life, your firm, your clients, or your business better?” He noted one potential he would like to see is the use of AI for employee reviews, with the ability to look over all communications with staff, their work and attendance records, their productivity, deadline management, etc., not just for firms, but for business management in general. This could provide a more objective overview of employee performance.
Blake Oliver hopes to see AI virtually eliminate accounts payable tasks, saving firms and businesses countless hours and headaches.
Nayo Carter-Gray’s ideal use of AI would be for healthy meal planning, with the ability to tell the system all the things she likes and doesn’t like to eat, while also keeping track of blood cholesterol levels and other health factors, with the system generating meal plans and recipes for her. It could then also create the InstaCart list and have the groceries delivered – and even map out the ideal use of the ingredients to reduce leftovers and food waste.
Hood added to the healthy meal and lifestyle AI possibilities by noting that there’s a benefit to having a non-human, non-judgmental computer telling you to watch your weight, diet, exercise, blood pressure or other health issues, as opposed to a person.
Tamera Loerzel stated she wants AI to figure out capacity planning, for both growth and projects, with automated scheduling.
Jenni Huotari wants AI to take over the work tasks she doesn’t like to do, to free up her time for the things she does like. This would include automating email and messaging responses, managing calendar items, follow-ups, deadline management, scheduling, and other background functions. On the personal side, she would love the technology to help manage family-oriented tasks.
Nicole Davis wants an AI assistant, “something that literally screams at me through my phone and says, ‘Hey Nicole, you’ve got your next meeting coming up. Stop what you’re doing and get to it.’ I literally need that in my life, because right now I ignore all the alerts in my phone and calendar.”
Final Thoughts on AI
The session wrapped up with a couple of thoughts on AI.
“My prediction is, within the next 5 years there will be a tremendous amount of $100 million, ten-person companies driven by AI,” said W. Michael Hsu. “We fundamentally have to rethink how business is set up.”
Kelly Phillips Erb had the closing thoughts, reflecting on how the legal profession has been impacted by some AI usage and rulings. “It’s imperative that firms and business leaders not drag their feet or try to prevent the use of AI in their organizations,” she said. “We’re already seeing ways that either customers or administrative agencies and judges are putting the brakes on ChatGPT use. If we don’t get out in front, some of these decisions will be made for us.”
How to tell if content was written by AI: https://techwiser.com/ways-to-check-if-the-text-is-ai-generated-by-chatgpt/
WSJ: Businesses Aim to Harness Generative AI to Shake Up Accounting, Finance
McKinsey: What Every CEO Should Know About Generative AI
Accounting Today: Large Accounting Firms See Promise in ChatGPT’s Future but Less in its Present
Accounting Today: Smaller Firms Curious and Cautious About ChatGPT
CPA Practice Advisor: Your Firm and Your AI Tools
Fortune: The way we work right now will soon look vestigial. Here’s how AI will make work more human
Asana CEO: ‘The way we work right now will soon look vestigial. Here’s how A.I. will make work more human’ |