How accountants are utilising AI in practice

aat comment

The Big 4 are investing in Artificial Intelligence, but is it worth it for smaller businesses?

The accounting sector is one of the largest adopters of Artificial Intelligence (AI) technologies. Indeed, the Institute of Analytics (IoA), the global not-for-profit professional body for analytics & data science believe that accountants are ‘ideally placed’ to plug the current AI skills gap.

Currently the Big 4 (Deloitte, PwC, Ernst & Young and KPMG) are leading the way with AI adoption by investing and trialling AI tools and software and utilising their capabilities and insights for business intelligence and data.

Interested in artificial intelligence?

Explore our collection of AI courses for accountants, from risk management to natural language processing.

Learn more

Despite AI’s capabilities, UK-based businesses in generally are adopting it relatively slowly. According to the 2022 report AI Activity in UK Businesses by Capital Economics for the department for Digital, Culture, Media and Sports (DCMS) just 15% of businesses are currently utilising AI to a certain degree, while just 2% are piloting AI technologies. An additional 10% are planning to adopt it in the future.

With the move towards automation and the increased need for accurate, real-time data, AI technologies such as predictive analytics and AI auditing can provide ‘greater insights’ to businesses and clients, Dr Clare Walsh, Director of Education at the IoA said recently.

AI’s potential uses in accounting include:

  • Predictive analysis (PA).
  • AI-enabled document reviews.
  • AI-enabled natural language processing (NLP).
  • AI-assisted forecasting and strategic planning.
  • Audit automation.

So how are smaller practices using AI? We found out which AI technologies accountants are currently trialling, and how professional practices are benefitting.

We use ChatGPT Google extension for routine client enquiries

Caroline Carter FMAAT, Owner, Carter Clear Accounting

We utilise a ChatGPT extension for Google that incorporates the capabilities of GPT-3, GPT-4, Bard, and Claude. This system is integral in identifying and responding to standard questions. Our inboxes are connected to our practice management software so we can use the extension to draft a response.

The technology makes our processes more efficient, especially when dealing with routine enquiries. It allows us to quickly and accurately address common issues, freeing up valuable time for more complex client needs.

However, one challenge we’ve encountered is the initial setup cost, which can be prohibitive for small practice owners. We have five users and this costs £120 per month. Security is also a concern, we turn off the extension until we want to use it – it’s designed to track activity across all web browsing. Additionally, accessibility to specialist training remains an issue. We had to learn in-house by trial and error!

In the past, we’ve also tried practice management software with built-in AI but it didn’t have the same functionality.

Looking forward, the future of AI in accounting holds potential for further streamlining processes, but addressing accessibility challenges will be pivotal for broader adoption in the next 5-10 years.

Verdict: At the moment we use ChatGPT Google extension to deal with routine client enquiries, but there’s potential for us to do more in the future.

We’re reviewing ‘use cases’ for document and template generation

Rob McGillen, Chief Innovation Officer, CBIZ Financial Services, Kreston Global

We’ve begun reviewing ‘use cases’ in document generation following templates and general communications enhancements. These have been exploratory for the past several months and we are shifting into demonstrable and reusable prompts.

Generative AI requires a bit of upskilling or training to find the optimal ‘way of work’ for individual professionals. We are providing prompt engineering training and demonstrations to help spark adoption within accountancy practices. Continuous engagement and training opportunities along with peer-to-peer knowledge sharing are sparking more adoption and interest than any other methods of engagement.

AI-related issues and challenges can be diminished with effective prompt building and custom instructions, as well as private data sets which have ‘walled garden’ limitations applied to the potential result sets. And not all platforms are equally effective at similar tasks – so finding the right tool for the task is key. As things continuously evolve (and quickly) that requires focus and updates to our cadre of professionals.

As a practitioner, generative AI is making certain aspects of the work more efficient because you are not focusing on ‘drafting and writing’ so we can spend more time on the insights and application of professional expertise. Generative AI (and AI in general) used effectively helps reduce lower impact utilization, allowing professionals to focus on their ‘top of license’ professional insights.

Verdict: We’ve been reviewing and trialling ‘use cases’ for document and template generation over the past couple of months, and are finding AI lets staff spend more time on insights and analysis.

We’re using AI to extract information and build internal performance frameworks but these tools aren’t quite as intuitive as they could be

Dominic Ahearn, Director, Best Suited Chartered Accountants

We are in the early stages of seeing what AI can be used for. Our practice management system, Karbon, has launched some AI features in beta, including summarising the content of emails in a helpful format. We also use Dext for processing purchase invoices and they use AI for extracting information. In addition, we use Chat GPT to help build out our internal performance frameworks.

We’re seeing some marginal wins in terms of time-saving and organising information. However, the tools available do not always feel as intuitive as they might and still require a lot of human effort to apply AI in a useful way.

I’d love to see AI being developed further for pulling information from different systems and creating client dashboards as well as providing instant insights tailored to an individual and firm’s way of thinking. Facilitating human absorption of information is really where AI can help us deliver our best work and most value.

Verdict: We’re in the early stages of using AI to extract information and build internal performance frameworks, but a lot of the technology isn’t there yet.

AI has huge potential but should never be used for making judgements

Aaron Westgate ACA, Chartered Accountant and Accountancy Teacher

Many accountancy practices are hesitant to use AI given its risks, so scope is often limited to less risky activities or internal processes.

Some software incorporates the use of AI (e.g. identifying risky transactions) but this is based on defined criteria so is more attuned to automation rather than AI. There have also been significant increases in data analytics within firms which facilitate machine learning (ML), and have the potential to provide significant added value or identify issues that humans wouldn’t have time to find.

Previously, small teams were firefighting and resolving data issues but there’s now more focus on data veracity and data source validation. We should now focus on investing in the advancement of ML and intelligent machines, hence we are seeing launches (or relaunches) of multiple intelligent engines, such as the infamous ChatGPT, and these are being integrated into more everyday tasks.

AI’s potential is huge but it brings a number of risks and questions. For example:

  • How transparent will firms be with clients about their use of AI?
  • How will AI affect fees and pricing models?
  • Will junior staff have the same level of understanding if ‘the system’ is doing it for them?
  • Staff will have to learn how AI works, not just what it produces.
  • Legal liability of AI will be up for debate as well as how it impacts professional indemnity insurance.

There’s also the assumption that AI knows what it’s doing. In reality, it’s an iterative learning process. AI should be limited to factual statements and assessments based on defined criteria, not making judgements that should be made by the firm. For example, whether a control within a business is effective, a sample size is appropriate or if a tax scheme is considered appropriate (and within the legal framework).

Verdict: AI has huge potential, but it’s an iterative learning process and we need to limit its use to factual statements and assessments for now.

If you would like to contribute to future articles like this one please get in touch with Annie Makoff-Clark at [email protected].

Interested in artificial intelligence?

Explore our collection of AI courses for accountants, from risk management to natural language processing.

Learn more

Annie Makoff is a freelance journalist and editor.

Related articles