Sunday, September 19, 2021

How AI Can Run Internal Audits

I watched the movie “The Accountant” recently, and it brought back memories of running a special tiger team internal audit effort at a tech enterprise.

The concept our then CFO had in creating the team was that rather than just doing compliance, we were missioned to look at business operations broadly, recommend staffing changes, and if we concluded an executive needed to be replaced, serve in that executive’s place until a replacement could be found. It was unique in both the company and industry. The company wasn’t into unique programs at that time, so even though we’d validated the program with impressive results, the team was eventually disbanded.  

But one of the audits that stuck with me was one where right after we left, fraud was discovered, and I took that result personally since we hadn’t discovered it. We pointed out that the poor controls in place could result in a crime like the one discovered, but I still felt we hadn’t done our job. The head of audit took me aside. I was distraught as I felt I’d let the team down, given I was the one who had personally done that part of the audit and pointed out that we tested by sample and the sample we’d pulled didn’t include the fraudulent work. Hence, the actual discovery was impossible, and we had identified the exposure.  

Let’s talk about how AI could fix this sample problem and better ensure financial processes: 

The Missing Link Of Digital Transformation

As we move more aggressively to digital transformation, such as in corporate finance, the missing link is how do we assure the process we are automating is a good process?

Digital transformation, particularly automating things, increases speed and efficiency, but if the things increasing are things that harm, then the process could make minor problems into far bigger ones.

This automated process validation would seem to be the missing link of digital transformation, assuring the various processes before increasing their speed by automating them.

What if, for instance, you didn’t have to audit by sample? What you could review the entire ledger and every transaction? An AI could be trained to do the tedious part of the audit job at scale.  

See more: Could AI Make Events Like CES Viable Again?

The AI in Audit

Much of what takes place in an internal audit is sampling reviews of the founding documents surrounding financial transactions. These records can be in the millions in a large company, and there is no way even a large team can review everything. So, a sample of the transactions is taken, and a review of the processes is done to ensure best accounting practices, like separation of duties and proper attribution of the financial elements, are in place. This review is tedious work, and since the sample is just a tiny part of the overall population, the crimes and mistakes that may be endemic to the process in place may not be identified. You certainly can identify their potential, but without evidence of an actual crime, management might choose, as they did in my case, to leave the flawed process in place to contain costs or simply because management doesn’t want to change.  

But AIs can review the content at machine speeds, and what the audit team is looking for in an audit are things like bogus vendors or payments that break trends. For instance, a large payment to a vendor out of line with past payments or for something that vendor doesn’t sell. Executives taking advantage of expense reimbursement programs aren’t uncommon either. For instance, we had a guy that would book airfare months in advance, then book it again right before the trip. After expensing it, he’d return the second ticket and fly on the earlier cheaper ticket pocketing the difference. Not at all an unusual scam. But you had to know about the potential as an auditor and then look for people who always seemed to book at the last minute, see if that was the nature of the job or an operation, and finally cross-check the airline records to see what the trip cost. All of this could be automated, and an AI could be trained to do all or part of this validating effort if it had access to the relevant data.  

More importantly, just knowing this was in place could prevent employees and executives from taking advantage, which could be an issue getting an AI project funded. If I were still in audit, I’d be taking a hard look at anyone blocking funding for a project like this.   

Wrapping Up

The protagonist in the movie “The Accountant” is a person with autism who is a numbers wizard and can perform a massive forensic audit at incredible speed due to his ability to focus and see patterns in numbers others missed. An AI, which can be trained to see patterns, is naturally far more capable than humans with numbers and repetitive tasks and could do what “the accountant” does in the movie, even faster and better than is portrayed.  

Using an AI to do entire population audits before the related process is digitized and automated should not only be more effective at catching and preventing fraud, it should also help find and correct processes that are doing harm. You better ensure that your digital transformation effort doesn’t do more harm than good by correcting those processes.  

I’m not the first to identify this missing link. AuditMap Technologies is working with Deloitte to create such a solution, according to Internal Audit 360. It would be worth checking them out.  

Similar articles

Latest Articles