Since the emergence of Chat GPT in November 2022, the global race to automate has gathered significant pace. But in certain sectors – particularly those involving complex processes and high levels of regulatory scrutiny – some market participants are questioning whether a degree of patience is required to mitigate risk.
In accountancy, a sector fraught with such risks, this debate is alive and well. Some firms are wasting no time in harnessing the technology to optimise service delivery, but others are taking a more prudent approach.
“Because[Chat GPT] is in open beta, I’m not sure if it has any formal place within accounting firms at the moment,” says Ben Bilsland, partner and technology sector expert at RSM, one of the UK’s leading audit, tax and consulting firms.
“One thing to be wary of is feeding privileged client information into it – that’s a problem for accounting firms.”
Bilsland goes on to clarify that these “short term risks” will eventually make way for “deeper and better insight”, but he ultimately argues that firms must remain vigilant in the meantime.
Part of this, he says, is simply reminding employees of the firm’s approved software platforms. He gives the example of Whatsapp – a messaging platform that, while commonly used, is an unapproved communications channel through which client information should not be shared.
At present, RSM currently views ChatGPT through the same lens, with the tool not being part of the firm’s approved tech stack.
‘Most firms will be using it’
Becky Shields, partner at mid-tier firm Moore Kingston Smith, believes the process of limiting Chat GPT use among the workforce may be less straightforward.
“You cannot fully control the tech stack anymore. Obviously, you can lock down applications, but with anything that’s web-based, the tech is there before you get the chance to acknowledge it.”
Shields compares the widespread and web-based nature of Chat GPT to that of Google, arguing that it would be “absurd” to limit accountants’ use of a search engine.
Firms need to consider that unless they have locked the use of Chat GPT on their systems, most will be using it “in one way or another”, she says.
“I think anyone who says their people aren’t using it is naïve. I cannot believe that your average graduate who has an interest in technology and understands the capability of this tool is not going to be using it in their day-to-day life.”
Adapting the strategy
Moore Kingston Smith is in the process of exploring the use of Chat GPT in every service line, with a view to rolling out a usage policy and training guide, says Shields.
During this process, the firm has discovered a potential use casein the production of research & development reports, where the science or engineering behind a claim has to be substantiated. “It’s really good at that”, Shields says.
“That’s one area where, as accountants, we employ specialists to do that, but we don’t have the ability to do it ourselves, and nor do our clients.”
She also cites code review as a potential use case, with Moore Kingston Smith being a provider of bespoke coding for audit analytics.
“For those sorts of really specialist scientific areas where you don’t have any capability, it’s hugely valuable.”
Big Four firm PwC has recently revealed its undertaking of a similar initiative. The accounting giant announced in March that it has partnered with Harvey, a Chat GPT powered AI platform.
The firm intends to integrate the tool into its tax and legal services arm, claiming that this will allow it to perform AI-led contract analysis, claims management, due diligence and other processes.
PwC will also look to develop and train its own AI models with in hopes of creating customised products and services.
This development of proprietary models, Shields says, will be a critical factor in the long-term implementation of language-based AI accounting firms.
“This is a very important focus of what the strategy should be. This bot has learned by being fed large volumes of information. So, we should equally invest as much time in categorising and cataloguing our own libraries and information.
“Get on board with what it can achieve and then think about how you can make it more relevant and more powerful internally.”
In spite of PwC’s AI-positive approach, others in the professional and financial services space have been less forthcoming.
London law firm Mischon de Reya, for instance, has warned staff against inputting client data into Chat GPT. Meanwhile, JP Morgan has restricted its traders from using the tool, with fellow banking players Goldman Sachs, Citi and Deutsche Bank reportedly set to follow.
Additionally, GCHQ, the UK government’s communications headquarters, has warned the public that Chat GPT and similar chatbots are an emerging security threat.
The education landscape
Echoing the views of Shields, Emma Rawson, technical officer at the Association for Taxation Technicians (ATT), argues that a proactive shift in strategy is necessary to adapt to the onset of Chat GPT in the education space.
She explains that, because exam cheating is alive on the radars of education providers like the ATT, the risks of generative AI being used in this area are being considered.
But rather than pushing back against the use of the technology, it is the responsibility of education providers to produce “well written exams that ask candidates to apply their knowledge, not just state the rules”, Rawson says.
“As it stands, Chat GPT shouldn’t be able to get around that as easily,” she adds.
According to Rawson, this comes down to allowing the benefits of the technology to flourish, rather than potentially suppressing them by pushing back against it.
“I do think there should be a certain attitude of how can this help us rather than whether it’s an existential risk to the industry. I’m very reluctant to ever push back against emerging tech because I think you lose opportunities by focusing on the risks sometimes.”