(AI Accounting – 5 minute read)
Admittedly, we’re all riding the AI Hype-train now, especially since Generative AI burst onto the scene in late 2022. If you say AI, chances are that you’re thinking of ChatGPT, Microsoft’s CoPilot or even Google’s Bard. But other forms of AI, notably machine learning and predictive AI, have been around for decades. Especially in accounting software (and spreadsheets!).
And guess what, in all the years that AI has been used in accounting programs, the demand for accountants has never disappeared. Let’s look at why the different types of AI will never replace the need for accountants.
Machine learning AI accounting
Machine learning is the term given to software that can recognise patterns in data. From an accounting perspective – this is good for forecasting large amounts of financial data. Machine learning software can analyse a large amount of financial data very quickly. Then it will forecast the future financial performance of the business using this historical data.
Examples of accounting software that use Machine Learning are Pocketsmith which can predict your future bank balances based on your household spending habits. For small business forecasting, Xero can provide your business with a short-term cash flow forecast based on your repeating invoices and regular monthly transactions.
Why machine learning won’t replace accountants
Unfortunately, machine learning is only as good as the data you put into it. As the saying goes ‘garbage in, garbage out’. Machine learning won’t give you an accurate picture if you only have a few months worth of financial data entered into your accounting system. Furthermore, if the quality of your data entry is poor, machine learning will spit out incorrect assumptions and bad forecasts.
Also, even when working with years of financial data, a lot of modern machine learning software still fail to get forecasts right. For example, Xero’s short term cash flow forecast severely underestimates your monthly cash inflow if you don’t have repeating invoices set up. So there’s still a fair amount of tinkering to do to get machine learning to reflect a more realistice financial forecast.
An accountant still needs to interpret the results of your machine learning software. Using their experience they can determine if the forecasts generated line up with their expectations of your business’ industry. They can also ‘sanity-check’ some of the assumptions made by your machine learning to ensure that forecasts line up with actual expectations. Accountants are also more in tune with movements in economy and industry, which allows them to tweak financial forecasts to be in line with actual economic happenings. A machine learning software trained on years of good economic times won’t be able to forecast bad economic times without human intervention.
So yes, machine learning is fantastic. But you’ll still need an accountant to interpret and make sense of the results for you.
Predictive AI Accounting
Much like machine learning, predictive AI also uses historical data to predict future outcomes. However, predictive AI looks at human user patterns instead of historical financial data. Almost all modern accounting software uses some form of predictive AI. If you’ve used Xero before, you will notice that once you categorise a particular type of transaction to an account code, Xero will eventually start recommending that you keep matching that transaction to that code.
Predictive AI is often limited to book keeping work. Some accounting software uses predictive AI to suggest words for you if you’re using the software to email a client an invoice. Most accounting software will also allow you to program ‘rules’ for the software to follow if a particular type of transaction appears. This will further enhance the predictive ability of the software.
Why predictive AI won’t replace accountants
This one’s easy to explain. Predictive AI is entirely based on human input. Therefore, if a human user has been incorrectly coding a transaction, the predictions generated will be incorrect as well.
A common example is car loan payments. A lot of business owners will code car loan payments as ‘motor vehicle expenses’. Do this often enough and predictive AI will keep suggesting ‘motor vehicle expenses’ when car loan payment transaction appears. However, car loan payments need to be coded towards the ‘car loan’ account (which containts the remaining balance of the loan). Car loans payments are repayments on the amount borrowed and cannot be claimed as a tax deductible expense.
Predictive AI won’t tell you if you’re doing something wrong. In fact, if you’re doing something wrong, predictive AI may give you false confidence that what you’re doing is right. We’ve had a lot of clients fall into this trap, which has resulted in a lot of work for our team to fix.
You will still need a bookkeeper or accountant to cross check your entries for you. Automating entries is a powerful tool. But you need to make sure that the initial inputs and rules are set up correctly. You can’t simply trust predictive AI to do all your book keeping work for you.
Generative AI Accounting
The poster boy of the AI hype – Generative AI! Tech giants like Microsoft and Google are all jumping on the Generative AI hype train. Accounting software have also been more aggressively marketing their AI capabilities since generative AI broke onto the scene.
However, generative AI has limited use in an accounting context. This is because generative AI generates content (like articles or drawings) and provides no financial insight whatsoever. That being said, you could feed a financial report to a generative AI (like ChatGPT) and ask it to summarise the financial report for you. But it won’t give you much more than that.
Why generative AI won’t replace accountants
I’ve experimented with the most recent ChatGPT 4.0’s ability to summarise P&L comparisons from PDF and I’m a little bit impressed. See below the output for a summary of a sample bi-monthly P&L:
So if you’re looking for a quick, top-level summary of financial information, it seems that generative AI does have some capability in that area. However, generative AI won’t explain to you the reasons why those movements happened. Yes, January was profitable compared to December. But WHY was it more profitable? An accountant will explain the reasons behind the movements. The savings in rent is due to a shift in premises. The profit increase in January was due to less money spent on stocks. December was unprofitable because a bulk order for goods for the next few months was made.
Generative AI can summarise financial information for you, but it won’t provide you with any insights. For that, you will still need an accountant to explain the numbers to you.
AI accounting won’t replace accountants
Granted, I’m open to the possibility that in the future, machine learning will provide better forecasts, predictive AI can spot errors before they’re entered and generative AI will get better at providing financial insights. But even then, the role of the accountant will evolve with technology, as it has for the centuries in which the accounting profession has been in existence.
Some accountants will get replaced though. Replaced by other accountants who are more tech and AI savvy. An accountant that can leverage technology to deliver quality outcomes to their clients with high efficiency will be highly sought after. Accountants who cling to the old ways will be left behind. But the accounting profession itself is here to stay. Business owners still want to talk to someone who can ease their financial anxiety and build their financial confidence.
So if you’ve got a tech-savvy accountant looking after you, good on you! And most of all,
Stay positive!