How small accounting businesses can use AI for big results

Early in 2023, KPMG Australia announced it was launching its own, private version of ChatGPT. It was being directly integrated into Microsoft Teams for all staff and was expected to “support the firm’s culture of innovation, boost efficiencies and create a better people experience,” KPMG Chief Digital Officer John Munnelly said.

Now, just about any firm can have this power – or something close to it. OpenAI’s latest ChatGPT allows paid subscribers to develop their own ‘GPTs’ on the platform, a new iteration of a product advanced users may previously have known as ‘wrappers’ .

With such advanced AI tools available to a great proportion of the global population, small accounting businesses must consider how they can use AI to augment their operations.

A 2023 study by the Thomson Reuters Institute, ChatGPT and Generative AI within Accounting Firms and Corporate Tax Departments, involving 771 tax professionals across the UK, US and Canada, concluded that tax professionals appreciated the potential of AI, but were still highly mindful of its risks. About 85 per cent of respondents came from SMEs.

One of the biggest issues, the study said, is the fact that ChatGPT and its ilk, having been developed by American companies and in a data environment dominated by the US market, has essentially been trained on American tax code.

Respondents to the May 2023 research also noted that the platform had only been trained on data up to 2021 – a moot point now, with paid subscribers able to use the tool to generate responses based on current online content, or even custom or curated knowledge hubs. This point reveals just how quickly any research or commentary on ChatGPT can age. So perhaps it’s worth reading the remainder of this discussion as ‘current as of publication’.

Right now, ChatGPT is able to answer simple tax questions but requires a robust fact-check. In its current form, it’s as valuable as an inexperienced accountant who requires close instruction and closer supervision. It can recite the basics but struggles with complexity. It can calculate debits and credits, and even identify trends, but isn’t client-facing.

Its ‘basics’ are still a powerful offering though. And its potential impact far surpasses that of a single graduate.

A study by Open AI (ChatGPT’s parent company) and University of Pennsylvania researchers concluded that more than half of their tasks of up to 49 per cent of workers could be exposed to Large Language Models (LLM) – the types of data sets on which ChatGPT’s knowledge is built.

The entirety of accountants’, tax preparers’ and auditors’ roles are exposed to LLM capabilities, the study said.

Will the robots take our jobs?

Does that mean robots will soon replace accountants? That’s unlikely, but robots will definitely change our jobs. And that is something accountants must prepare for.

“The onset of AI, I think, is worrying for the profession, in that clients and prospective clients may not see the need to engage with a professional person and simply ask the question online,” says Ian Hornsey FFA FAIA, Chairman and Board Director of the IFA, and Joint CEO of Devonports LAS Accountants.

“Therefore, the accountancy profession must encourage regulators and governments to monitor this carefully. AI is potentially a tool for accountants to use, albeit carefully.”

AI as a tool will likely arrive in the form of customised software, developed to empower accountants to do higher-value work.

“Our analysis suggests that, with access to an LLM, about 15 per cent of all worker tasks in the US could be completed significantly faster at the same level of quality,” the OpenAI/University of Pennsylvania study concluded.

“When incorporating software and tooling built on top of LLMs, this share increases to between 47 and 56 per cent of all tasks. This finding implies that LLM-powered software will have a substantial effect on scaling the economic impacts of the underlying models.”

This indicates the scale of AI tools’ potential but, right now at least, it really is just potential – it still only augments the human user. The Thomson Reuters Institute study concluded that while ChatGPT’s answers aren’t always correct – “or worse, they may be just accurate enough to mislead people” – expectations are that its performance will improve.

The most serious concerns are around confidentiality, accuracy, privacy and security.

“Tax and audit professionals have been here before, of course — with electronic calculators, keypunch machines, Excel spreadsheets, TurboTax, tax automation, and other ingenuities that were supposed to hasten dramatic change within the profession,” the Thomson Reuters report said.

“And now it is time once again to assess the promise and peril of a new technology that almost everyone has heard of, some are already using, and few truly understand.”

About 75 per cent of respondents to the study said they believed AI could be applied to tax, audit and other accounting work, but only around 50 per cent thought it should be.

At the time of the survey, in May, just over 10 per cent of respondents said they were using AI. But 51 per cent thought their businesses would roll out some sort of AI-powered tool in the next six to 12 months. Six months have now passed.

Are you using AI in your practice? We want to hear how you are using AI tools, what you have learnt along the way, and what you’re planning next. Get in touch.

Small accounting businesses and AI

In terms of use cases, most respondents suggested AI for tax research. Other popular perceptions of the use of AI included compliance, back-office functions, tax return preparation and as a question-answering service.

Benefits overwhelmingly centred on the time-saving elements of AI usage, and the fact that it would liberate people to do higher-value work.

Businesses using ChatGPT for more creative requirements that weren’t so reliant on security, confidentiality and accuracy, such as requesting a checklist of useful processes for onboarding a client, were more satisfied with the usefulness of AI as a workplace tool.

Thomson Reuters Tax & Accounting says the four main use cases in small accounting businesses right now are:

• Bookkeeping automation such as account reconciliation and expense categorisation
• Client communication such as chatbots for simple client inquiries
• Data analysis and back office functions such as the analysis of client data
• Tax research which must still be edited by a knowledgeable human

What software is available right now?

Some interesting AI offerings have already been developed for accountants in smaller businesses.

Of course, ChatGPT is being used for everything from writing email copy to designing details and schema of specific business databases.

A subscription-based, America-centric app called ChatCPA, made specifically for accountants, works within an individual’s favourite messenger app and answers questions on accounting matters.

An AI accounting app called Digits, marketed as the “world’s first AI accountant”, claims to manage books in a live environment, providing an executive summary, P&L statement, payroll expenses, balance sheet and cash flow in real time.

Dext automates bookkeeping and expense management, offering insights along the way via accounting practice data. And Futrli uses AI to forecast cash flow and other reports up to three years in advance by using financial data to predict real outcomes.

Clearly, AI is on its way into accounting practices of all sizes. How those practices choose to use the technology remains to be seen.