With the increasing immersion of generative AI into the workplace, Franki Hackett, Head of Data Ethics at Engine B and Sarah Schlobohm, Data and AI Leader, discuss the skills accountants need to use it effectively.

How is generative AI being used already?

As in many other professions, there are accountants are starting to draft some of their reports using generative artificial intelligence (AI) tools, such as ChatGPT or Bard. While this work obviously needs human checking, this basic time saving is one way this technology is already having an impact.

“I have heard of people using it to create the first draft of their annual report, and particularly the notes to the accounts, because they are so structured, which is quite an easy thing for a generative AI or word-based generative AI to do,” says Franki Hackett, Head of Data Ethics, Engine B and chair of ICAEW’s Data Analytics Community.

Sarah Schlobohm, Data and AI Leader, says that auditors are already using generative AI for a range of tasks, including the code for models and forecasting. She says: “You’re going to need more auditors with the skills and ability to understand these models and the risks that are introduced by companies using these models.”

According to Schlobohm, auditors need to ask whether companies are using models within frameworks properly, and if government guidelines are being followed when looking at the risks and not just the factual errors.

Accountants not only need to be aware of generative AI, but they also need some understanding of how it works, so that they can use it correctly and be aware of how it could be misused by others.

Hackett warns: “I don’t think accountants are thinking about how other people are using generative AI to produce things that might not necessarily be reliable, but that they have to engage with.

“As a former auditor, the big scary version of this is people using generative AI to produce fake invoices, fake transactions and fake evidence that is very hard to tell apart from real examples.”

Hackett, who is speaking about the impact of technology and audit at ICAEW’s Annual Conference in November, predicts greater use of things like circularisation confirming balances and transactions to try and manage this risk.

What skills do you need to use generative AI?

Schlobohm believes that the most important first step for accountants is to try using generative AI now and get a good understanding of the technology while the barriers to entry are low.

“Absolutely everyone needs a general awareness of generative AI, how these things work and what the big risks are you need to think about when you are using them,” she says.

Hackett suggests that a lot of the skills accountants need to work with this technology are ones they already have. “We are good at looking at and reviewing evidence. We are good at synthesising information. We are good at balancing information and making judgement-based decisions,” she says. “What is needed is developing confidence and competence in understanding AI systems and how they work.”

While there has been a lot of talk about prompt engineering, Schlobohm argues that this language isn’t something to get hung up on “Prompt engineering is a fancy way of saying ‘ask good questions’ and accountants and auditors are trained to ask good questions in a professionally sceptical way,” Schlobohm says.

One key skill that accountants need for any kind of technology is a basic understanding of data structures, confirms Hackett. This means understanding the difference between string data (which is alpha numeric characters you might think of it as language) and numbers (numeric data) is really important.

“Understanding that data is often stored in tables and how those tables relate to each other is also important,” says Hackett. “Understanding, for example, how the general ledger relates to the trial balance in a data sense, and in a systems sense, is really going to be increasingly key.”

Helping to boost your confidence

The best way to feel more confident about this technology and its potential impacts is to get familiar with it, according to Hackett. “The more widely you read, the more information you will get and the more confident you will feel,” she says.

“There are brilliant sources online, such as Gartner, but often very simply googling something is worthwhile. You will find a lot of basic answers that will explain concepts in lay language and will refer you onto other things.”

Schlobohm agrees: “There are lots of good free, or not terribly expensive, resources that are available. There’s been some good stuff released recently from Google, Andrew Ng and DataCamp. And always check out the government guidance, as we know laws and regulations are going to have a huge impact on the industry.”

ICAEW has a range of resources to support members in getting to grips with generative AI including the AI hub on icaew.com, which brings together the latest articles, webinars and reports.

There’s also the Data Analytics Community and the ICAEW Annual Conference this autumn, which is themed on technology and has a keynote on the impact of generative AI on the accountancy profession.

Ian Pay, ICAEW’s Head of Data Analytics and Tech, says “Accountants are always being encouraged to be curious, and nowhere is that more apt than when exploring exciting new technologies like generative AI. Never before has a single technology generated such a buzz globally - we can't be left behind.

"It's great that at the ICAEW Annual Conference later in the year we'll have a chance to hear directly from Microsoft and others about the potential of this technology in our sector.”

For accountants interested in this technology Hackett recommends being brave. “Don’t be too nervous to use it, don’t lock down and say ‘the regulators are never going to like it’ and it’s not worth trying. Don’t think it’s too scary... Have a go, what’s the worst that can happen?”

Schlobohm agrees that getting involved early on is a good idea, but advises caution when it comes to confidential information and warns that the technology isn’t a replacement for human intelligence. “Don’t use it exclusively for anything where the answer has to be 100% absolutely correct. It’s great to get a first draft, but always read it through carefully,” she says. “And always test it if you’re doing something technical.”