Computers falling in space
From Pivot Magazine

The AI accounting challenge: how to turn it to your advantage

While there’s still a lot to worry about when it comes to AI, accounting and financial professionals are finding ways to make it work for them

Computers falling in spaceSince the 2010s, when digital assistants like Siri and Alexa hit the market, technological leaps have thrust several AI tools into everyday life (Photo illustration by Cindy Lubinic)

Bryant Ramdoo, partner and National Audit & Assurance Innovation Leader at KPMG in Canada, began his career as a CPA in 2006. At the time, Facebook had just launched, Apple was filing trademarks for its now iconic iPhone, and Ramdoo was still using paper audit files in binders. Suffice it to say, artificial intelligence was in its infancy. The concept of advanced AI was mostly relegated to studies like those in scholarly articles at the University of Toronto (U of T). In 2006, Geoffrey Hinton, a University of Toronto faculty member, cognitive psychologist and computer scientist, along with a Ph.D. student, Ruslan Salakhutdinov, had just published a research paper on how large neural networks—the bedrock of deep learning algorithms—could be effective in real-world applications.

Fast forward to the 2010s, when digital assistants like Apple’s Siri and Amazon’s Alexa hit the market. Since then, technological leaps in computer hardware and software have thrust several AI tools into everyday life. “Now we’re at the point where there’s been an explosion of data, along with tools and technology that allows us to process that data,” says Ramdoo. “Companies are now harnessing that data to conduct financial audits with the help of AI.”

Bryant RamdooBryant Ramdoo, national audit and assurance innovation leader, KPMG Canada (Photograph provided)

While these advances have exciting applications for fields like accounting, researchers and experts are sounding the alarm on the dangers associated with the potential misuse of AI. Even Hinton, who is regarded as the Godfather of AI, stepped down from his role at Google in April 2023 to speak out on the risks of AI. Hinton joined Google to conduct research in deep learning in 2013 while remaining with U of T’s Department of Computer Science. Still, experts in accounting and AI agree that if executed carefully, these powerful algorithms can make the jobs of CPAs easier.

Take auditing, for example. As any auditor will tell you, the job involves time-consuming, repetitive steps, from poring over sample transactions to analyzing data sets to conducting surveys. This is where companies like MindBridge come in. Based in Ottawa, MindBridge is a risk discovery and anomaly detection platform that uses AI to help auditors make their jobs more effective by identifying, surfacing and analyzing risk across broad financial datasets. The company has raised over $40 million since its founding in 2015 and recently announced a global alliance with KPMG in April 2023. KPMG is infusing the power of MindBridge AI into its smart audit platform, KPMG Clara, to improve anomaly detection in digital audits. Where auditors would traditionally only be able to find a needle in a haystack of risky financial transactions, Ramdoo says MindBridge has already looked at every piece of hay so it can be identified in seconds. “So, instead of trying to find the needle in the haystack, the tool gives us the ability to look at an entire field of haystacks at once,” adds Ramdoo.

Rachel Kirkham, VP of AI and Product at MindBridge, explains it can be difficult for auditors to figure out where to focus. “If you are a large company with lots of data, pinpointing the areas of risk is hard,” she says. MindBridge can process up to 500 million records for an analysis, and then analyze that data to identify the transactions that need further investigation. “In an external audit, that gives you a better understanding from the starting point about what the risk is in the data,” explains Kirkham, who spent nearly 10 years at the U.K. National Audit Office researching the applications of AI and leading a team of data scientists and analysts in developing data analytics capabilities for financial audit. “It allows auditors to focus on sampling or selecting transactions that look most unusual to execute their audit testing. So this is much more efficient than the traditional methods.”

MindBridge uses different risk indicators—machine learning-driven, statistical or rules-based—and then combines them into an ensemble AI, which allows its users to focus on the most unusual transactions. Those transactions could be found in a company’s general ledger, payroll, charge card or a list of vendor payments, Kirkham says. The software loads the data, runs machine learning analytics, and then the output shows the user ranked risk information, allowing the auditor to see risky transactions more easily. “When I started out, I was the guy who had to assess those individual transactions manually,” says Ramdoo. “Now we have a tool that, with a couple clicks, we can get insights that would’ve taken weeks or months if we even tried to attempt that historically.”

Besides saving time, Ramdoo explains the MindBridge tool embedded in KPMG Clara allows auditors to understand parts of algorithms and exactly why that transaction was triggered as high risk. “It’s avoiding situations where an algorithm just picks a transaction, and we can’t explain why,” he says. This is known as explainability—a set of processes and methods that explains results to human users in a way they can trust and understand.

Geoffrey HintonGeoffrey Hinton, cognitive psychologist and computer scientist, was once known as the godfather of AI. But now he warns of the existential risks of the rapidly evolving technology (Photograph by Chloe Ellingson/The New York Times/Redux)

The first step to transitioning to using AI in accounting and finance, Kirkham explains, is to get familiar with the different types of AI and how they apply to accounting. MindBridge, for example, looks at transactional information and uses a type of AI known as anomaly detection, which is different from something like a large language model, which forms the basis of tools like OpenAI’s ChatGPT.

Once a company has a good sense of how AI tools will help it and how those tools will fit into the company’s data strategy, then MindBridge’s Kirkham suggests thinking about what change management strategies will need to be put into place: “Does it require people to think about change on a managerial level, whether it’s people, processes, or tech?” Based on those decisions, companies can then think about the processes that need to be changed and what skills their CPAs (present and future) need to use them.

It can also help if there’s someone already at your firm who has a deep interest in the technology. This is because a new AI tool is kind of like an intern who’s eager to learn and contribute to the company, says Erin Kelly, president and CEO of Advanced Symbolics Inc., an AI-focused market research firm based in Ottawa. When training an AI on bookkeeping, users have to start “thinking” like the tool so they can figure out the best ways to get it to do things properly. “Now, if you get the wrong person to do it who doesn’t have that enthusiasm, they could end up cutting corners,” she says. “And when you cut corners, the AI gets stuck on that, and it will compound the errors.”

With that said, the problem really lies in security and privacy, says Kelly, referring to experts like Hinton who are voicing concerns about the dangers of AI. “We need new legislation, and we need to get started now because AI technology is coming out the door faster than the government is dealing with it.”

Since accounting looks different in every company, incorporating AI is more than a one-size-fits-all solution, Kelly emphasizes. She recommends firms consider how they plan to use the technology, especially since sensitive financial data is involved.

CPA involvement is important, says Kelly, in order to avoid handing over the metaphorical keys to an AI tool. We’ve had autopilot and AI on airplanes for a long time, but flights don’t operate without a pilot on board. “It’s the same thing for accounting,” says Kelly. "You still have to be involved, you still have to watch it and make sure it’s not making errors. And you, a human person, have to make sure you catch those mistakes. You don’t want to get to the end of your fiscal year and find out that the AI was constantly being mixed up on something.” Programs for upskilling CPAs on new tech are available in Canada. For instance, KPMG Digital Academy was built in partnership with Simon Fraser University in Burnaby, B.C., to prepare KPMG auditors for a future in accounting with courses that centre on data analytics, innovation and AI.

Some of that training even starts at the university level. At U of T, accounting professor Minlei Ye assures that the curriculum has evolved to ensure a higher level of technological proficiency. This includes courses that cover topics like data analytics, basic coding and the computer programming language Python. “I think the curriculum changes at the university is adapting to the changes in the broader auditing industry,” says Ye.

Ye sees the potential for AI to be used in other areas of accounting like applying AI tools to help companies determine what accounting principles or policies to use in deciding the allowance for doubtful accounts. Typically, you would use past experience or estimates to decide how much percentage of the revenue will be a write off to an allowance for doubtful accounts, explains Ye, adding that an AI tool could make that estimation more accurate, saving businesses both time and resources.

On the soft skill side, Irene Wiecek, a professor of accounting at U of T, thinks it’s important students and CPAs be open to change, including technological changes. “It’s being open to all the changes that are coming now and will continue to come,” she says. “In the competency map, we called it evergreen learning—and it’s a willingness to continue to learn and also a commitment to continue to learn.” Wiecek was one of eight members on the Competency Map Task Force, which finished its work in March 2022. While the new competency map, CM 2.0, does not explicitly require universities to teach students about AI, Wiecek explains it’s much more high level. Rather than having a list of topics that must be taught to CPA students, it leaves some flexibility to meet the needs of the university’s employer partners that are hiring graduates. “We do talk about the CPA having to be tech savvy,” says Wiecek. “So, emerging technology mindset and evergreen mindset are absolutely central to CM 2.0.”

“If you want a career as a CPA, I think there’s a lot to be optimistic about, such as not doing some of the mundane things and harnessing these tools to help you do your job better,” says Ramdoo. This, he says, can look like spending more time driving insights for your client’s business, doing better analysis and having better—human-to-human—conversations.


Check out CPA Canada’s extensive tech resources to find out how automation and AI could change the CPA’s role. Read our Q+A with industry expert Olivier Blais on whether AI is moving too fast for regulators to keep up. Listen to the Foresight podcast on the impact of AI on the accounting profession, and discover the importance of a human-centred and responsible approach to AI.