As it has with almost every industry, technological advancement is transforming the realm of accounting. Today’s auditors are relying on cutting-edge tools, such as artificial intelligence (AI) and machine learning to help them sift through vast quantities of data and pinpoint any anomalies for their human counterparts to analyze.
Naveen Kalia is an Audit partner with over 20 years of experience with KPMG in Canada’s financial services practice. He is also the Canadian firm’s Partner In Charge of Audit Innovation, responsible for identifying and integrating new technology and tools that are changing the face of auditing.
Eli Fathi is the co-founder and chief executive officer of Ottawa-based MindBridge Ai. The company’s premier product is Ai Auditor, an audit tool that harnesses machine learning and artificial intelligence to help human auditors deliver a higher-quality, more accurate audit.
Given the rapid changes in the industry, we asked Kalia and Fathi to discuss the question: What does the audit of the future look like? Their answers have been edited and condensed.
How has the auditing landscape changed from 10 years ago?
Kalia: We have far more access to our client’s data than we have ever had. You couple that with new auditing tools to analyze that data - be it robotic process automation, or AI, and you have a completely different way to look at an audit. All these changes are possible because of our access to data. Big Data.
Fathi: Now you’re dealing with two sets of data: You have the structured data – the numbers. But you also have unstructured data, such as e-mails, invoices and memos exchanged between companies, for example.
How has all this data changed the approach to auditing?
Kalia: Historically, our audit approach was all sampling-based. You would analyze those samples and you’d come up with an extrapolated, statistically valid audit result. Now, with our ability to process large amounts of data, you have the ability to look at everything.
An audit of the future lets us run all our clients’ data through our audit tools and then allows us to focus more on the anomalies and the outliers. This allows us to do more complex risk assessments than we were able to do in the past, resulting in a higher quality, more focused audit.
What new tools are having the greatest impact on auditing?
Kalia: The first big development was the advent of tools that allowed us to capture all our clients’ data and have it scrubbed and useable – essentially ensuring it’s complete and accurate. We’re now in the second phase, which takes that data and tests it against a series of rules. This rule-based automation flags anomalies and any entries that don’t make a lot of sense, so we can analyze them in a focused manner.
AI is what I would call phase 3. Now that we’ve programmed in all these rules, can we program a machine to learn and modify these rules as it goes and apply its own judgement and how we as auditors control that?
Fathi: In our Ai Auditor system, we have over 80 algorithms. The most exciting is called reinforced learning.
With reinforced learning, the auditor teaches the system. If the auditor says, ‘This is an error,’ we record that, and as we get many of those over time, the system becomes an expert and learns how to do it by itself. We expect that [eventually], the system will be so smart that it will be able to handle the whole assessment without really needing human intervention.
But we don’t foresee that in the next three to five years. It is augmenting what the auditors are doing, rather than displacing them.
How does this new model change the skills an auditor needs?
Fathi: The skills you need are the ability to comprehend, to be flexible and to be innovative.
Kalia: We will still need the traditional auditor skill set when it comes to understanding judgements made in financial statements, following up on identified issues and in dealing with the client. I do however think their skills needs to be tweaked a little bit.
I always tell students, ‘You should be taking data courses, not learning how to code, but learning how to interpret results.’ At KPMG in Canada, we’ve joined forces with the Beedie School of Business at Simon Fraser University [in Vancouver, B.C.] to give our people the opportunity to get a Master of Science in Accounting with Digital and Cognitive Analytics. The firm will pick up the cost for that because we understand that these skills will be essential to serving our clients now and in the future
What about the EQ side of it – the emotional intelligence required to work with clients?
Kalia: It’s still critical - and it could be even more important and a differentiator going forward. As with any industry, clients will always want to work with someone they find trustworthy, reliable and savvy. As auditors, we often have difficult conversations with our clients. They’ve made a judgment and sometimes we’ll come up with a different judgment. That’s not going to change whether it’s AI or whether it’s human.
In the end, we are accountable for explaining our judgement. We can’t just tell a client the machine made the decision. We need to understand what is behind it and then sit down with the client to explain our reasoning. If you can’t do that, you’re not going to have clients for very long.
Fathi: As an auditor, you’re dealing with the chief financial officer, the audit committee and the CEO. The numbers are the numbers, but when you have an incident, you need the emotional side. Those are the soft skills that auditors in the future are going to need more and more.
Are auditors resistant to these changes?
Fathi: It’s the tale of two worlds. Organizations are traditionally structured as pyramids. At the top, you have the experienced people and, at the bottom, you have the newcomers. Today, if you look at most of the CPA firms, they’re no longer a triangle shape, but rather a diamond shape.
Why? Because millennials don’t want to be just number crunchers. They are looking for something that is more fulfilling and gives them purpose. AI will help fix that because we’re going to remove the boring, labour-intensive part of the job and allow them to do more meaningful work like risk assessment based on the data collected from the machines.
Kalia: I tell new students that I would be excited if I were them – more so than nervous. Rule-based automation has allowed us to focus on the riskier transactions, where in the past it was more mundane sampling. And then when we add AI to the mix, understanding the judgments becomes even more interesting. It’s more fun as an auditor to follow up on things that don’t make sense, compared to just sampling.
What are the biggest challenges of using AI in audit?
Kalia: With any technology you are always worried about privacy issues, with AI specifically you get concerned around bias in AI.
The challenge is: How do you test [a system] in a robust-enough fashion that you’ve taken the bias out of the machine? And that really comes down to volumes and volumes of data that you have to put through the machine and test over and over again.
That is the biggest jump to get to phase 3 [and true AI]. Phase 2 is straightforward. I’ve programmed in a bunch of rules. I know exactly what I’m going to get out from it. Whereas in phase 3, the machine is starting to create its own rules. It comes down to, can I understand what the machine has done? Can I explain it, can I replicate it and can I prove it?
Fathi: From day one, we decided that our AI would be explainable. When we show you a result, we actually demonstrate to you in plain English how we derived it. You can actually go and see which algorithm fired and why. So it’s just not a ‘black box.’
People are concerned that the algorithms will be biased and are going to create the wrong results. The way that we have overcome that is we are not using a single model, but we are using what is known as ensemble AI. Due to the number of algorithms being used, it eliminates the bias.
Privacy is a major, major thing.
Then there’s also the ethical issue. In December, [Canadian AI pioneer] Yoshua Bengio [and the University of Montreal] launched the Montreal Declaration for Responsible Development of Artificial Intelligence. We were the first company that signed it, aside from the not-for-profits.
In essence, the question is: Who is judging the judges? Governments around the world now have to come up with an independent body that can verify that algorithms are doing what they’re supposed to do.
What innovations are you most excited about?
Fathi: Blockchain is exciting because it will basically guarantee the providence of the data at the point of origin. A blockchain is a distributed ledger that authenticates transactions between two or more individuals.
But let’s not forget, virtual reality and augmented reality are going to be extremely important. Imagine that in 10 or 15 years, auditors will wear goggles and see the data floating in the air. That’s going to be amazing technology for sure.
Kalia: To me, the future is the real-time audit.
If we look on a long enough horizon, we can get to a point where I push a button on Jan. 1, and I produce a set of statements that have been audited, because I’ve done a real-time audit throughout the entire year essentially being plugged in to our clients system, auditing transactions as they happen (perhaps through a blockchain).
To me, that would be the coolest thing that we could do, and I’m hoping I get to experience it before I retire!
Advertising feature produced by Globe Content Studio. The Globe’s editorial department was not involved.