(This column originally appeared in Forbes)
For accountants like me, the promise of artificial intelligence is exciting. Why wouldn’t it be?
The profession’s biggest firms – like EY and PwC – are deploying AI technology in their auditing and financial review procedures in order to identify irregular transactions or patterns of inconsistency. The technology could ultimately be used to perform a much deeper level of auditing analysis that can offer both investors and financial statement readers a more accurate and complete picture of a company’s finances and at a more affordable cost.
Accounting software provider SAP recently released its AI assistant called Joule which promises – like Microsoft Copilot and similar offerings from Salesforce – to “provide customers with contextualized information and assistance when performing tasks in SAP applications.” SAP is already building an AI foundation layer with Joule sitting on top in order to help customers do things like create job descriptions and interview questions as well as analyze financial data, forecast sales trends and help with decision making.
In a keynote speech this past fall, Oracle founder and CTO Larry Ellison said that his company’s cloud capabilities will be leveraging AI to resolve customer service issues faster and – in the healthcare field for example – help providers reduce their manual work through voice commands.
ERP maker NetSuite is adding AI capabilities throughout its financial products too. The company has recently added its own generative AI application called NetSuite Text Enhance for use across HR, finance, supply chain, and sales. The tool promises to generate “content for any text area based on a few starter words that describe intent.” According to the company “NetSuite Text Enhance helps finance and accounting, HR, supply chain and operations, sales and marketing, and customer support teams improve productivity by leveraging AI to produce relevant drafts that they can quickly and easily review, edit, and approve.”
These are just a few examples of accounting products and platforms that are just starting to leverage generative AI. Other major providers from Sage to Epicor to Microsoft Dynamics, Xero and QuickBooks are doing similar, exciting things. But take note of the words in all of these statements, press releases and news articles: they are “could” and “will” and “building” and “promises.” It’s all just starting. Accountants should be aware: none of this stuff is ready yet for prime time.
Why? For three reasons.
The first is articulated well by Rajiv Rao, a contributing Writer to ZDNet, who warns that if you want to use AI tools for accounting, “you better think again.” Rao cites a recent survey that pitted ChatGPT against humans to solve accounting problems. Unfortunately, AI didn’t do too well, with students scoring an average of almost 77 percent correct answers vs. ChatGPT’s 47 percent success rate. “According to the study,” Rao wrote. “The LLM (Large Language Model) did fine on things like auditing, but had trouble getting its artificial neurons around problems that dealt with tax, financial, and managerial assessment problems – and these were sections that involved a lot of math. The output of these AI LLMs hinges on probability, and not accuracy. Output, by design, has been architected to represent an answer that has the statistically highest probability for the question asked. And numbers, sadly, don’t work like that.”
So AI really has a ways to go before it’s figuring out accounting problems.
Which is why you’re not seeing a lot of AI in the technical capabilities of these accounting applications. Even with all the hubris around the “transformative impact” of AI, what’s telling to me is that these new AI “features” really aren’t featured very much in these companies products. Research firm Gartner, who publishes their quarterly “Magic Quadrant” which looks at the leading cloud enterprise resource planning (ERP) systems in the world, barely mentions AI features in the major accounting platforms in its latest report. Why? Because there really aren’t any, at least none that would make a difference to the accountants using these products.
Finally, even the software vendors themselves are telling the world that their AI offerings aren’t really there yet. Per the sources I’ve reference above, few details of PwC’s AI technology have been disclosed with the firm’s spokespeople admitting that “it feels like something we should be developing or exploring”, and that early results showed the technology has “legs” for auditing. Analysts of SAP say that Joule “has promise” and that rolling it out will “take time.” And although NetSuite touted its AI tools in its recent press release the fact is you have to read a little before you realize that most of the features added to the product’s new iterations have little do with AI.
This doesn’t mean that AI won’t live up to its promise. It’s just that the AI functionality teased by the largest accounting and ERP vendors is still being developed and we have a ways to go – maybe a few years even – before it gets rolled out, perfected and incorporated into our companies’ and clients’ financial processes to the extent that we can actually rely on marrying this automation with our data.
For now, accountants and financial professionals should be paying attention and getting ready. We need to prioritize cleaning up our databases, going to vendor conferences, demanding demonstrations and understanding exactly how these tools will help us do things quicker and more affordably. That way when prime time comes, we’ll be good and ready.
