AI for accountants: science fiction or fact?
30 October 2020: With artificial intelligence promising to be a gamechanger for accountancy, what’s the right approach for accountants looking to incorporate AI into the mix? Mark Blayney Stuart investigates.
Perceived wisdom states that artificial intelligence (AI) can help accountants in two significant ways – it can automate basic but time-consuming tasks and then, as a consequence of this, free up accountants’ time to do more sophisticated advisory work. But if, according to Sage, 58% of accountants believe AI will help their business, why is uptake not as fast as it could be?
The root of the issue perhaps lies in the mix of uncertainty around what the technology actually is, does, and whether it will be worth the investment. Clearly, AI is here to stay. And an indication that the technology has already fundamentally changed the nature of what the accounting profession is can be seen by the $9bn being spent by three of the Big Four on AI, data analytics and training staff to use advanced technologies.
So, how can you decide what is right for you? “I think whether a piece of tech is generally useful is a matter of individual preference,” says Robert Collings, Finance Lead at Flux. “For me, you need to focus on either things you dislike doing or things that take too much time.”
Adopt this principle and, Collings says, you are in a clearer position to make decisions. “An example of this for me is expense processing,” he explains. “I’ve been in a role where expense claims were completed via Excel, then printed and handed into the finance team each month. It’s painful and time-consuming, so moving to an employee spending solution, such as Pleo or Soldo, was a game changer.”
However, not every solution necessarily resolves a problem. “I was recently looking for a platform to manage departmental budgets and tried a few providers,” says Collings. “Too many of them were either too complex or just too difficult to understand. And that’s me as a finance person saying it. Who knows what would happen when department heads log in! In that instance, the tech is just not worth it.”
But what about COVID-19? Does Collings think the pandemic has made some technologies more attractive or necessary than others? “One of the AI developments we’re seeing a lot at the moment is around cashflow forecasting. If an app knows your historic spend and near-term payments, based on due dates, it can do a reasonable job of forecasting your future cashflow.”
However, as with many other current solutions, the need for humans becomes apparent when things get a bit more complicated. “It works for the simplest of businesses where there isn’t much change month-on-month. The moment you throw in an edge case or a changing environment, it usually can’t cope,” he adds.
This is partly because, as Collings explains: “It only sees a limited dataset – your accounting system – and doesn’t yet have the capability to search for external clues.” But what if AI were to have the ability to read emails or listen to phone calls in the future? “As intrusive as that sounds, perhaps it could then build a more predictive model that knows if you’re about to buy a new car, raise investment or write off a huge debtor?” says Collings.
The more AI can discover independently, the more it can be helpful. But the privacy implications of that are significant. Ultimately, it will be an individual choice about how much your AI assistant knows about you.
Two further advantages of AI outlined in the ICAEW Thought Leadership report, Artificial intelligence and the future of accountancy, include improving fraud detection through more sophisticated, machine learning models of ‘normal’ activities and better prediction of fraudulent activities, and improving access to, and analysis of, unstructured data, such as contracts and emails, through deep learning models.
And for strategic business and technology adviser Bernard Marr, AI can “provide real-time status of financial matters since it can process documents using natural language processing and computer vision”. This advance, in Marr’s opinion, makes “daily reporting possible and inexpensive”, which in turn lets businesses be “proactive and adjust course if the data show unfavourable trends”.
Marr takes up the fraud protection point by suggesting AI can help “support auditing and ensure compliance by being able to monitor documents against rules and laws and flag those with issues”. AI technology is able to do this because it can “quickly sift through enormous amounts of data to discern potential fraud issues or suspicious activity that might have been otherwise missed by humans and flag it for further review”.
On the one hand, AI has been promised as a panacea for industry problems and on the other a chimaera that keeps failing to deliver on its promise. An entertaining but slightly sobering example of this is the Botkeeper saga, described as an AI bookkeeper replacement, which in reality turned out to require offshore humans to complete month-end accruals or deferrals, administer payroll, pay employees and bills, project cash flow and complete bank reconciliations.
Cutting through the hype
With that in mind then, what can AI realistically do next? “I’d love to see a more connected finance world,” Collings says. “Why do I still need to export things from one system into another? Why do I still need to receive a PDF invoice from someone when invoices are so standardised? Why do I still need to make payments manually when the invoice contains all the details needed for a system to set up and schedule the payment itself? Having a more connected finance world would make things a lot easier!
“I’d love to see AI really take on those edge cases and have the ability to not only prevent errors but also to fix them. Things like picking the correct nominal code based on a range of data points. It would get to the point where it knows its own limitations and then you can manage it by exception – rather than randomly spotting errors.”
Ultimately, AI’s incremental developments will continue to speed up the working day, help with process efficiency and enable the move towards business advisory. Yet, as the Botkeeper example shows, there will always be parts of the profession that require human input and analysis. The robots are not going to take over just yet.
Mark Blayney Stuart is former Head of Research at the Chartered Institute of Marketing.