KPMG and MindBridge are partnering to integrate machine learning technology and AI technology into KPMG’s Clara, the company’s global audit platform.
The alliance with MindBridge, which will embed AI on audits across the KPMG global network, will bring increased levels of AI into Clara, which can leverage better insights and a greater understanding of complex sets of data.
The integration of MindBridge technology should also allow for improved risk identification, helping in the continued delivery of higher-quality audits by combining KPMG’s industry experience and MindBridge’s techniques.
These make use of MindBridge’s advanced statistical, machine learning, and rules-based analytics technology.
“KPMG auditors will benefit from using MindBridge’s AI-embedded audit intelligence tools, visualized analytics, and the in-depth resources needed for stronger analysis and assessment of risk,” says Larry Bradley, global head of audit at KPMG. “That means spending less time on routine data reviews and instead will increase focus on identified relevant riskier items.”
He explains as part of this strategic alliance, KPMG audit professionals will have increased access to no-code AI capabilities and learnings, designed to enhance their ability to analyze high volumes of data to garner increased insights and value.
“With this new lens, KPMG auditors can see new insights into clients’ business that drive better analysis, better conversations, and better quality,” Bradley says.
Lou Trebino, audit chief technology officer for KPMG U.S., points out auditors are always adapting to changes in the macroeconomic landscape.
“Technology disruption, geopolitical risks, recessions, and more,” he says. “The past few years have certainly been disrupted and highlighted the importance of technology in helping auditors’ effort to continually enhance audit quality.”
In the past, audit teams working to analyze millions of revenue transactions had to use sampling to generalize about all the transactions, based on insights from a subset of them.
“With AI, we can analyze large, complete populations of data and flag anomalies for human review,” Trebino explains. “That process allows auditors to zero in on potentially risky transactions and do further analysis. It’s a much more efficient way to find possible risks and stay ahead of the curve.”
He adds unlike popular discussion of AI, the model also identifies why the outlier was flagged–it’s not a “black box” for the auditor.
“If the technology flags an outlier that the auditor finds was actually recorded appropriately, that new information is incorporated into the process via machine learning so that future analyses are more accurate,” he says.
From Trebino’s perspective, the increased capability to analyze entire data sets and find outliers is a game-changing next step in our digital transformation.
“Using automation frees up auditors to do work that requires more judgment, which in turn provides a better audit for the client,” he says. “Importantly, clients are using AI and building it into their own internal controls.”
As they develop their own modern approach that lines up with the audit profession, they are receiving greater insight and can communicate with their auditors more clearly, quickly and accurately during the audit process.
“It also will demand that auditors learn how to audit AI with AI,” he notes. “Moving ahead with our own AI capabilities will help us lead the profession into a future where AI will help us with audits of financial data but also AI technology itself.”
Billy Cheung, solutions architect at governance software provider Diligent, notes it’s important to note the difference between generative AI and predictive AI.
“Generative AI, which is used in tools like ChatGPT, can be a helpful research and admin tool,” he explains. “The value lies in time savings and consistency of document quality. It can suggest areas of focus that you might not have considered, provide a different perspective or summarize a detailed technical analysis into a more consumable format.”
However, any information provided to a public AI can and will be used by the AI for others, which means that must be very thoughtful and measured in using this capability appropriately and safely.
“In contrast, predictive AI is leveraged most commonly for audit, and uses techniques such as machine learning to detect patterns, classify and categorize data and assist in decision-making,” Cheung says.
While random sampling relies on luck and the biases of the sampler, a combination of predictive AI and traditional rule-based testing allows us to review entire populations on both trends and behaviors (predictive AI) and internal guidelines (rules-based testing), giving a more comprehensive and higher level of assurance.
“More importantly, how the information was analyzed and why deviances were identified will also become more transparent, vastly improving the defensibility of the audit results,” he says.
From the perspective of Phil Lim, director of product management at Diligent, the world is only just now beginning to unlock the possibilities of AI, and audit is no different.
“A key challenge for using AI in audit is how easily we can interpret and explain the outcomes,” he says.
In other words, when organizations use predictive AI to detect patterns, they must understand and be able to defend the reasoning behind how it detected those patterns, along with any assumptions and biases implicit in it.
“This is the typical concern noted around deep learning,” Lim says. “As explainable AIs continue to develop and these concerns are addressed, the use of AI will become more integrated with common audit methodology guidelines.”
