AI in Accounting: Five Key Points from AICPA ENGAGE

minute read

Last Updated July 6, 2023

Category Tech trends


Artificial intelligence won’t replace you…a person using AI will!

That was the message that permeated throughout presentations I attended at this year’s AICPA ENGAGE Conference. Without a doubt, AI in accounting encompasses the most significant new block of technologies to hit our profession (and frankly ALL professions) in the last two decades. The consensus is that it will have a major impact regardless of whether you like it or not!

As background, the definition of AI is highly individualized by the different exposures that each of us has had to the technology. This, of course, results in extremely varied perceptions and reactions to the concept. These perceptions range from the “super intelligent computer personas” popularized in movies at one end of the spectrum to vendors touting applications that “use AI” to perform fairly mundane tasks. For accountants, those tasks include identifying accounting anomalies, scanning and populating tax returns, and pre-coding expense reports.

AI in Accounting Is Still in a Discovery Stage

I believe the reality behind the hype surrounding solutions such as OpenAI’s ChatGPT, Google Bard and Jasper Chat is somewhere in the middle of what has appeared in the mainstream press. To be fair, ChatGPT is just one of many AI tools available today. However, it is garnering most of the attention because it was the first “consumer”-grade tool available and is really easy to use.

Within the accounting profession, most professionals seem to use ChatGPT to pull together marketing research and content…and summarize information. Jason Staats and Hector Garcia have popularized some great use-case examples of accounting data analysis coding with ChatGPT. But those use cases hardly fall into the category of an easy-to-use, turnkey experience for the majority of today’s accountants.

Why AI in Accounting Can Lead to Security Risks

There are a few critical tips to keep in mind. We should never submit client or confidential data into these public chatbots. This careless practice has led to multiple instances where users inadvertently exposed confidential data. Firms should immediately educate their personnel on what these products do and make sure they understand why they should never enter confidential information into them.

Doing a little reading on the recent Samsung ChatGPT leak will help you eloquently summarize the risks that led to very strict usage rules and education across corporate America. Companies such as Microsoft, Apple and Amazon (which banned usage to write or debug any code Amazon may use/develop) implemented policies around using ChatGPT.

AI in Accounting Is Becoming Easier to Use

A second focal point is that AI in accounting will see faster adoption when it is easier and more intuitive to use. It will really take off when it runs in the background with our applications and data…and, of course, with the right security and our trust. We’ll know we trust AI when we’re as confident and comfortable in using it as we are with Excel or QuickBooks®. Intuit recently announced integrated Generative AI via its GenOS offering.

With Generative AI, we are seeing their first steps towards a consumer-grade AI solution for targeted use with the accounting data inside of QuickBooks Online (QBO). Imagine typing in, “Show me my 10 largest client receivables balances that are over by 60 days, and provide me the primary contact’s name and phone number for each,” and getting a listing. This is where the accountant’s knowledge in asking questions gets a big assist from AI.

Specialized Tools Are the Future of AI in Accounting

A third takeaway was that smaller, specialized GPT tools trained on very precise data are much more accurate and efficient than the massive language models that ChatGPT and Bard were trained on. When it comes to tax data and recommendations, the information has to be absolutely right or there could be significant financial consequences. I see this as the onramp for the major tax application vendors or even a very entrepreneurial tax expert to develop specific AI tools.

Those tools could provide very specific knowledge and solutions to complex tax questions while still protecting the specific client data or scenario. That way, accountants could use the tools to verify and document information with real citations. These AI assistants would rely only on trained information and would NOT create random new data when a valid answer does not exist. They would also incorporate specific technology to ensure that data stays within the firm’s “guardrails” without being exposed to any outside parties.

AI in Accounting Is Still Prone to Deliver Misinformation

One real potential pitfall for now is that in the absence of trained data, Generative AI systems will completely fabricate information that emulates the “very authoritative” resources you were expecting to get as a response. These are referred to as “hallucinations,” and all of us can get a first-hand example by asking ChatGPT to prepare a “detailed career biography about [your name here].”

In my instance, it had me graduating from a school I never went to. That sounds like no big deal, but there have been some incredible hallucinations. The now infamous ChatGPT lawyer airline case demonstrates how ChatGPT created six completely made-up legal court filings, complete with opinions and citations. Unfortunately, a lawyer submitted under the belief that ChatGPT was a search engine! Users must independently verify ALL references cited by AI chatbots.

Accountants Need to Develop Specific AI Skills

One final point to keep in mind is that these “artificial assistant” tools are already here. I believe “consumer-grade” accounting versions that make them easy for any of us to use are on the way. So, we should start getting familiar with how to use them. Learning to ask specific questions to get the information you want is called “prompting.” We need to use our accounting knowledge and skills to ask the right questions of our accounting data.

When we do, we can help ourselves and our clients make better decisions and discover opportunities. This is an advisory skill that most accountants have developed through years of working in the profession. We will have to learn to phrase those questions so that our AI assistant knows how to respond properly.

Understanding and bringing together these key points in your practice will help you become a person using artificial intelligence rather than being the person being replaced by it in the future.

Be ready for whatever comes next by running applications in the cloud. Start today.

Recommended next

Why You Shouldn’t Fear Artificial Intelligence in Accounting

Subscribe to our blog

Get Rightworks articles delivered straight to your inbox.