Through Section 2, we have looked at the trends in and some actual case studies of the use of AI in the accounting and finance domain, as well as the prospects of the technology in the future. In this section, we will look to consider the division of roles between accounting/finance personnel and AI, assuming that the application of AI will take a further step up going forward.
Firstly, we want to think about what the fundamental differences between AI and people are. We believe the differences between AI and people lie in whether they are able to “make judgments” and “be creative.” AI is a so-called “tool,” with its standout feature being that, through repeated learning, its accuracy can be improved. For example, take the case of improving the accuracy with which AI reads documents, or the case responses drafted by a generative AI. In either case, the machine is doing no more than just responding based on results learned from past information. Consequently, “judgment” and “decision making” about the validity of AI responses and the validity of outputs generated by AI can only be done by people, no matter how far the technology goes.
“Creativity” works the same way. AI, which excels at learning past information and outputting the right answer based on what it has learned, cannot express “creativity” in the true sense of the word, producing wholly new ideas that have not previously existed.
Given the above, the big differences between AI and people are “judgment” and “creativity.” We believe this is the key to thinking about the division of roles between AI and people.
Now let’s look to apply this to actual accounting and finance work. For example, applying this to the process of receiving an invoice and issuing payment delivery voucher, the invoice sent by the customer is read with AI-OCR, then AI determines the transaction category and automatically issues a voucher for the journal entry. Up to this point is AI’s role. Subsequently, an effective division of roles would involve people determining the reliability of the voucher that was read in and directing it to be recorded in the accounts.
The process of detecting fraud would work similarly, with AI detecting that a voucher that differs from the norm had been recorded. A person would need to ultimately determine if the details of the transaction were really fraudulent, and then take appropriate action. It would also be possible for AI to make sales forecasts and projections of final results, and display multiple patterns for these based on past trends, with it being up to people to determine the reliability of this work.
So far, we have spoken about “judgment,” but the same extends to “creativity.” Say a CEO orders cost reductions. When working on considering what specific cost reduction measures to take, an AI could perform various assistant operations, but it would be up to people to do the thinking. The “creativity” portion of coming up with what measures to take is something we believe it would be up to people to determine.
Given the above case studies, the division of roles between accounting/finance personnel and AI in the era of AI is likely to grow ever more significant, with AI likely to take on rote work such as issuing vouchers, aggregating and processing data to output reports, and collecting necessary data, while people will have an even greater role in determining the validity of vouchers issued, of determining whether the information gathered by AI is true, and deciding the reliability of content of information output by AI.