The main problem to address if you want to adopt AI effectively

Accountants are adopting Generative AI and are optimistic about applying it. Here are practical ways to use it.

AI is well and truly here. 91% of accountants are using artificial intelligence (AI), or plan to use it in the future, according to recent research from Wolters Kluwer. That figure is – unsurprisingly – higher among 18-24-year-olds than accountants aged 55 or above, but the overall data shows the prevalence of the technology in the accounting sector industry in the UK. And it finds that generative AI is the most popular type, used by 70% of accountants.

Peter Stiff is Head of AI and CPD at Astranti Group, a technology training and consultancy company that works with businesses to develop better AI skills in the workforce. He says the penetration of AI into everyday operations is already well underway, with a growing number of practices waking up to the potential benefits and overcoming their initial scepticism. 

“I think even up until a few months ago even when it was really hyped, people were still very dismissive; whereas now accountants are starting to realize that it’s not a fad or a flash in the pan – it is here to stay,” he says.

That’s reflected by a growing body of evidence that accountants are growing more curious and enthusiastic about AI’s potential benefits:

• 64% say AI will benefit their productivity
• 51% are sure it will save costs
• 38% hail its impact on automating workflows.

Indeed, AAT’s research has found that two in five people would consider a change of career to accountancy if they could use AI to replace routine tasks. 

The main problem for practices to address

However, even as acceptance and curiosity is growing, Stiff says there remains one problem in particular that practices need to address.  

“I think the key mistake that people make is not having a clear goal for their AI implementation.  

“It’s very easy to just say, ‘Here are some shiny tools like Chat GPT or an app that creates presentations or videos, or does your auditing and bank reconciliations’; but you should be asking, what do you actually use it for, and what do you want to achieve from it?”  

Simply adding in AI tools with the idea of “using it where you can” does not, Stiff suggests, get you very far.   

Instead, “You want to work backwards with the idea: this is what I want to achieve, this is my objective from implementing AI; and then work from there and begin to ask yourself how you are going to pilot it, test it, refine it and get people using it effectively.”

Safe sandbox

Unheralded benefits of AI are beginning to emerge from initial scepticism and concern over bias and accuracy.

“I think using AI has a huge benefit for relationships, as counterintuitive as that might sound, because, even in a simple model like a GPT, you can build what is essentially a digital twin of your client company.  

“To do that, you put in all the information and create specific GPT, put in all the branding and instructions relating to that company, all their materials. By doing that you can literally analyze your company’s numbers in real time because you’ve got their up-to-date systems in there you’ve got a pre-programmed GPT that is indexed to that company’s history and information.  

“Doing that provides a safe way of developing a training tool for new juniors. So it acts in the same way as simulations that allow traders to trade with fake money.”

Importance of validation cannot be overstated

AI-driven tools have already come to market to help with specific accounting areas such as the month-end close, tax and auditing. But AI is no silver bullet.

As more accountants take the plunge with a whole range of shiny new tools, they’re finding the output of generative AI in particular can be biased, inconsistent and inaccurate.  

Perhaps the most obvious risk with AI is the reliability of its results. However, 67% of accountants say they usually or always trust AI (according to the Wolters Kluwer report), so we cannot overstate the importance of validation.

“I think not validating the output and blindly trusting what is produced is one of the biggest risks of AI,” says Stiff. “The difficulty is that these tools – Gemini, ChatGPT, Claude, etc – have a great way of sounding incredibly confident.

“Sometimes the mistake will not even be obvious, it will be subtle. Only an expert will be able to spot an error easily.”

This may appear to pour cold water on the promise of AI, but really it should reassure accountants that AI is not an existential threat to the profession. “Accountants need to realise that their jobs aren’t going to be taken by AI,” Stiff says, pointing out that clients and accounts will always need external expert verification, especially when it comes to sensitive financial information and projections. 

Changing skillset

There will be changes, but subtle ones.

“Previously, accountants may have done say 50% of the work, 25% of the set up and 25% of the checking, with accounting software doing the rest. With AI, accountants will likely do 20% of the work, but their initial input and final validation will be indispensable.

“We will still need accountants, but the role will change as will the skills required,” he concludes.

Christian Doherty is a business journalist and freelance writer for AAT.