AI in Action

At the December 2019 Committee on Finance and IT (CFIT) Meeting in Dallas, Tx, I attended an amazing presentation called “AI - Powered Financial Audit.” It was an informative presentation by FEI partner named Trintech and their vendor MindBridge, The Mindbridge Artificial Intelligence (AI) product enables auditors to uncover unknowns in their data and to make decisions with more confidence  I have a professional and personal interest in learning as much as I can about AI.

Although I do not work for an auditing firm, it was helpful for me to learn how auditors can use AI as tool and view an AI demo in action. Machine Learning, Domain Expertise and Statistical Methods make up the AI tool. As the AI Auditor product runs through a client database, these database reviews are a form of machine learning. The tool scans your company database seeking, scoring and detecting outliers in typical reports. Data plays a crucial role for auditors and advisors with AI detection, or finding the needle in a haystack, AI prediction, or understanding complex relationships and Data visualization, or making complex information digestible. The visualization makes it easier to communication AI derived information to a client. It also allows clients to more easily identify problems and growth opportunities, make more informed decisions and monitor the progress that has been made.

Machine learning algorithms can understand complex relationships much better. For example, for an unusual cash disbursements rules-based system, a transaction is flagged as normal putting it in the 30th percentile of risk, however, for machine learning, the same transaction flagged in the 3rd percentile of risk. Machine learning is proving a better tool to understand risk factors and produces better outcomes than rules-based approaches alone.  

Data integrity, fraud frameworks and fraud reviews were part of the process. The best way to gain assurance is by looking at the relationships between data points. This needs to be done within a dataset and across data sets. The statistical modeling included Benford’s Law, regression analysis and three-digit testing.

AI will help the auditor to focus on specific issues and help to verify the presence of controls and good accounting practices. The presenter reviewed two sample use cases, one a consumer products manufacturing company and the other a CPA firm, where AI was used. The use cases illustrated how AI was beneficial to the audit workflow of risk assessment and planning, risk response and audit evidence and audit conclusion and reporting. My favorite part of the presentation was to observe the predictive analytics in real time. I have read about the process, but to see it build a visualization of a forecast model was very interesting.