Haven’t got time to read this story? Listen to it in audio format.

AI is different from almost every other technology implemented in accounting. There are risks around transparency, privacy and bias, and increased reliance on AI gives the technology more authority and an agency of its own. With AI in a first wave of implementation at scale, what are the ethics around communicating with clients on how it is being used?

AI tools, says Kayur Patel FCA, are like “having the world’s smartest junior at your disposal for 30 dollars a month”. Accountants who use AI tools are able to “utilise an army of digital brains”, which improves the productivity of their practice.

Patel, a director of PwC New Zealand’s gen AI and emerging technology team, has around 80 AI assistants and agents he uses regularly for his own work. He uses only subscription-based, enterprise-level versions.

“There is literally nothing I do now which I don’t do hand in hand with AI,” he says. “AI does the first draft of everything, but I still own the output. It’s my name and my signature that goes on the deliverable when we give it to a client.”

“If I’m giving clients an output, that means it’s my level of quality and that of the firm.”

Sharing knowledge

More and more clients, says Patel, are also using AI. In his role, he has worked with close to 200 clients over the last 18 months, helping them with AI “in some shape or form” in their own businesses. When it comes to the work that Patel does, clients now expect that AI has some input in its preparation.

“There’s no pushback, but some clients are interested and want to know more about how we use it,” he says. “They ask about our policies and if we are using it safely. So, we have all the documentation on that and we can provide it to clients, and we certainly don’t hide it.”

Patel doubts, however, that every organisation will make a decision on telling clients “this output was used in AI, this one wasn’t”.

“To compare, we don’t disclose to our clients when we’ve used our best junior or our worst junior on a job,” he says. “Ultimately, if I’m giving them an output, that means that it’s my level of quality and that of the firm. We stand behind that, regardless of the tools we have used.

“The important thing is that once we deliver it, we stand by it and own it.”

Open and transparent

AI’s use in professional services is an emerging focus at the Cranlana Centre for Ethical Leadership at Monash University in Melbourne. Cranlana chief executive Matt Finnis and program director Matt Beard make the point that accounting already has a strong foundation in a code of ethics, which deals extensively with issues of conflict of interest, transparency, privacy and disclosure, so AI fits into a long-established ethical framework.

Within this framework, Beard says that the “primary currency” of the professional relationship is trust. This requires accountants to be well prepared to answer client questions on issues such as where data lives on platforms, how those platforms are built, and their accuracy and reliability.

Just as the wider population is challenged by financial literacy, many people and organisations on the client side also struggle to understand technology.

“There is a mounting challenge around technological literacy and the way in which AI functions, and being able to understand the details,” he says. “What we are seeing with AI is a gap and an asymmetry between who is responsible for asking the ethical questions and who has the responsibility, and answering questions on whether the ethics are reflected in the design and if it delivers protection.”

There is danger in accountants moving too fast in using AI, then “taking the client out of the dialogue and assuming they don’t really need to know about it”.

“That is when accountants will start to run too far ahead, and they are going to leave their audience behind,” he says. “We are not at the point where everyone is so familiar with AI, that it is in common usage and it is a reasonable assumption that clients understand how this technology is being used.

“The idea that you can assume that ‘everyone knows that we’ll be using AI for this, so I don’t really need to have a conversation’ is probably too self-serving.”

This means that accountants should “always err on the side of disclosure”.

Finnis says that accountants need to keep remembering that “AI is going to continue to make mistakes”.

“The risk is that the humans won’t notice the mistakes,” he says. “Clients might not expect that your advice is perfect, but they expect honesty in every relationship on every occasion, so that means that transparency is critical to the equation and that extends to using AI.

“You can delegate as much of your work to AI as you like, but you can’t abrogate your professional responsibilities.

Case study: MGI+MORE

At MGI+MORE in New Zealand, chief executive Dan Henderson says his firm began using robot process automation (RPA) some time ago, and is now using AI for a range of functions and processes. The firm’s use of AI is ultimately good for clients because it gives accountants back time which can be re-invested in the client relationship.

With 10 offices and a headcount of about 300 across New Zealand, Henderson says the firm has focused on practical, high-impact cases such as streamlining reporting and dashboard creation, summarising complex email trails, analysing and comparing documents, and helping teams structure and interpret client data.

“More often than not, we use AI to generate the first cut of an analysis or scenario,” he explains. “It might outline the structure, compare trends or identify the variables worth paying attention to. Our accountants then come in behind that to test the logic, challenge assumptions and add professional judgement.”

MGI+MORE’s client base is largely family-owned and privately held businesses, and they are all made fully aware of the firm’s use of AI.

“If there’s an issue, we just don’t use it but 80% of the clients are totally fine with the concept of what we are doing and why,” he says. “We also talk to our clients about how they can be using both RPA and AI in their own organisations.”

Take away

Through the CA Library there is a LinkedIn Learning video, ‘Introduction to AI Ethics and Safety’, that explores potential biases, hidden consequences and risks of using AI models; discusses how to identify biased data sets and models; and includes case studies.