There’s no shortage of advice on how to use AI to boost efficiency across your business.
From LinkedIn prompts promising to automate entire accounting systems to increasingly sophisticated AI tools built into everyday software, businesses have more ways than ever to hand over financial tasks to technology.
But when the work involves sensitive data, regulatory obligations and decisions that directly affect a business’s bottom line, there’s plenty to consider before pressing go.
Andrew Erkins, founder and CEO of outsourced finance and bookkeeping firm Digit, says the challenge is knowing where AI can genuinely add value to your finance function without introducing unnecessary risk.
“You can’t just implement an AI system and then let it go,” he tells SmartCompany.
Smarter business news. Straight to your inbox.
For startup founders, small businesses and leaders. Build sharper instincts and better strategy by learning from Australia’s smartest business minds. Sign up for free.
By continuing, you agree to our Terms & Conditions and Privacy Policy.
Erkins takes us through the biggest watchouts when introducing AI into financial workflows.
1. Start with the problem, not the AI hype
Businesses risk getting caught up in AI for AI’s sake, rather than starting with a clear problem they’re trying to solve.
“If you’re automating stuff that doesn’t need to be automated, is there value in it?” Erkins asks.
For businesses looking to introduce AI into their financial processes, Erkins recommends starting small and focusing on tasks that are high-volume, repetitive and relatively low-risk.
“So if an AI system makes a call, it’s easy to undo that call. It’s not a terminal problem for the business if it gets it wrong,” he explains.
Instead of building an entirely new system from scratch, Erkins suggests businesses dip their toes in the water by exploring the automation capabilities available within the software they already use, such as Xero or MYOB.
2. Don’t let automation come at the expense of control
Greater efficiency should never mean less accountability.
Erkins says the three areas where human oversight is most critical are those involving money, legal obligations and decisions that directly impact people.
“Anything where money is going out the door, whether that’s bank details, or bank details that are changing for a supplier or for a customer,” he shares. “You want to confirm that is correct. You don’t want to automate that.”
The same applies to compliance obligations, such as BAS lodgements and financial reporting, where inaccuracies can lead to legal consequences.
Rather than reducing internal controls as processes become automated, Erkins argues that businesses should strengthen them: “Don’t take away controls. You need to make sure you’re adding controls to protect against or to know that the information is correct.”
3. Be clear about which AI tools make the cut
AI adoption can move faster than company policy, particularly when employees are left to decide for themselves which tools to use and what information to share.
To avoid leaving these decisions to individual discretion, Erkins stresses the importance of establishing clear guidelines around approved tools and use cases before rolling out AI across the business.
“One of the dangers is that if you don’t be really clear with your team about what tools are approved, then people will make their own call and will try things without thinking through implications around data protection and security,” he says.
Known as shadow AI, the use of unauthorised AI tools can introduce risks that businesses may not even be aware they’re taking.
“You don’t want to stick someone’s licence number, address, phone number, date of birth and mother’s maiden name into an AI system,” he says.
Erkins also encourages businesses to look beyond general-purpose AI platforms like ChatGPT and Claude and explore more specialised AI tools designed for financial workflows.
4. Make sure your data is AI-ready
Before introducing AI into your financial processes, take a closer look at the quality of the data you’re working with.
Erkins points to the example of a client who enabled automated bank reconciliation without first addressing inconsistencies in their existing records.
Rather than fixing the underlying problems, automation simply reproduced them at scale.
“It takes a mess and just amplifies the mess,” says Erkins.
5. Don’t underestimate the value of expert advice
Handing financial tasks to AI doesn’t remove the need for financial expertise; it increases the value of people who can interrogate the outputs and understand their implications.
“The risk for businesses wanting to automate their finance functions is really in not understanding what they need to check in order to know that the results are correct,” says Erkins.
Even when an automated reconciliation or financial report appears accurate, businesses need someone with the expertise to identify errors, question unusual results and understand the implications.
“You still need the humans that understand the topic deeply enough to know how to train and guide the system so that the system does things well,” he adds.
“And that’s our role: being the experts who can help train and guide AI systems to manage the finance function.”