By Charis Thomas, Chief Product Officer at Aqilla
Ethical use of AI in an accounting context means using the technology in a way that maintains the human components of authority and compliance when producing accounts. That means accountants must not abdicate responsibility for accuracy or accountability.
As accountants, we should be confident that we can vouch for the validity of the data that’s produced—and can verify AI’s thinking and conclusions, should we be called on to do so. For me, ethical use of AI means putting checks and balances in place to verify the data it provides and ensuring we don’t passively and unequivocally accept the results. We should use AI to speed up the process of making predictions, not to deliver assertions.
Ethical concerns of the use of AI in accounting
Since the boom of generative AI at the end of 2022, there have been concerns across the economy about types of AI usage. When it comes to the accounting sector, I believe the top three can be summarised as:
- Over-reliance on AI’s judgement and surrendering human checks and balances
It’s often tempting to accept all AI-generated content and reporting at face value without carrying out some manual checks. But we must not abdicate responsibility. Saying, “The AI told me,” is not good enough. We should be able to refer to the raw data source, put that raw data into spreadsheets, and produce the reports ourselves—albeit far more slowly. We must always be able to fully verify the integrity of anything we ask AI to produce.
- Blind trust in AI’s version of “the truth”
We’ve seen AI-generated pictures of people with two right hands, six fingers or one leg. That’s because when ChatGPT or other open-source AI can’t find a piece of data, it takes a best guess or makes something up. It often leads to amusing images, but they should make us uneasy and remind us how AI operates when it doesn’t have all the data it needs for a particular task. Imagine if that logic were applied to producing a set of accounts!
It’s also worth remembering that AI’s definition of the truth is only as accurate as the information it’s provided—that’s actually a profoundly human characteristic, too, when you think about it. If everyone were to tell ChatGPT that the sea is orange and made of jelly, it would present that statement as the truth. Again, imagine how this way of operating could impact the creation of accounts or the validity of financial data. That’s why, when we set AI a task, we must always be able to verify and understand each step that it takes to produce our data. We should also be able to verify that data at each point along the way.
- Ignoring the potential risks of AI-assisted shortcuts
It might be tempting to use unsanctioned open-source AI when short on time, up against tight deadlines or to improve efficiency. This can expose an organisation’s data to public access—and might open up wider IT security vulnerabilities, break data privacy laws and contravene industry regulations. Is that shortcut worth it, or should you put in the hours and do the manual work? The speed at which AI can run calculations might make us think it’s worth the trade-off, even when we know deep down that it’s not the right thing to do.
Finding the balance
If ethical concerns around AI use are addressed and remain front-of-mind, accounting teams can and should take advantage of the technology’s many benefits. They could use AI to run checks on transactions and accounts to gather data at a faster rate than any human could, or to assess compliance with industry-specific and general UK laws. Using the data and information it has gathered, AI can identify patterns, conduct trend analysis, and advise on organisational risk, all of which adds extra value to the business.
AI can also deliver more advanced automation of core accounting processes, such as capturing invoices and allocating suppliers. Passing these tasks over to AI frees accountants from the monotonous data entry tasks, whilst eliminating the possibility of human error, saving time and derisking processes. Once trust has been built with AI systems, finance teams can go a step further and even teach it to pay invoices—most likely those below a certain amount or to pre-agreed providers to reduce risk. Although these might sound like small individual tasks in isolation, the time savings are impressive when multiplied across an accountant’s entire workload.
Of course, this still needs to be balanced to address the ethical concerns—you wouldn’t want to let AI loose on your systems without any supervision. But running alongside manual human checks to ensure quality and accuracy, AI can transform an accountant’s workload and job satisfaction. Ultimately, it frees accountants up to do what they do best: analyse and critique financial data against corporate strategy and add real value to their clients.
Ethical AI in practice
Our ethics shouldn’t change when we adopt AI. If we have strong personal and organisational procedures, processes, and principles, they should already provide most of the guidance. However, given AI’s transformational capacity, we have an ethical obligation to educate ourselves about the technology. So, my advice would be to:
- Understand AI capabilities through personal use before engaging with the technology in a commercial setting
- Implement AI gradually, starting with easily verifiable areas
- Treat AI as a predictive tool rather than a definitive source
- Have clear sight of the data sources and decision-making processes AI uses to generate reports
- Balance AI assistance with human oversight and maintain responsibility for its output
I recommend using these pointers to build out a framework and guardrails for ethical AI use. This will ensure that everyone using the tools understands the basics and avoids innocent but potentially damaging, costly, and regrettable mistakes.
