KPMG — which has been experimenting with and deploying aspects of artificial intelligence for years across its business — is taking another step forward to boost its audit platform with AI, Sebastian Stöckle, global head of audit innovation for the Big Four firm told CFO Dive.
The company is deepening its ties to the Ottawa, Canada-based MindBridge, announcing a new alliance late last month that will bring the tech firm’s statistical, machine learning and rules-based analytics technology into KPMG Clara, an existing digital audit platform.
The news of the alliance comes as ChatGPT, an AI-driven natural language processing tool created by Microsoft-backed OpenAI, has generated significant buzz — critics say hype — about its potential for supercharging the digital transformation of business. It’s also drawn backlash stemming from concerns about threats the new technology tool poses to jobs and its vulnerability to manipulation.
But in an interview Stöckle said that the advances in AI that the alliance is embedding in KPMG’s audit process are aimed at enhancing audit quality and efficiency, by reducing errors and in some cases identifying fraud. He emphasized that people will still be in control of the process. “It’s always with a human-machine interaction,” Stöckle told CFO Dive. “It’s never a ‘leave the machine alone’ type of scenario.”
The Berlin, Germany-based Stöckle has been with KPMG for about 17 years, during which he has audited and advised clients around ERP systems and focused on data analytics as it relates to audits. Most recently he’s been globally responsible for developing audit analytics at the firm, working for Larry Bradley, KPMG’s global head of audit. In that role he’s been very focused on developing technologies like analytics, automation, blockchain and AI, he said.
“At KPMG we have been experimenting, developing and deploying different levels of AI over the years for different purposes in audit, tax and advisory,” he said.
While generative AI models like ChatGPT are dominating the public discussions of AI right now, Stöckle said AI is actually a very broad term that has multiple levels and variants.They range from the entry level AI called machine learning to a subset known as unsupervised machine learning, to deep learning. Even the more advanced language models now in the headlines are just “the tip of the iceberg,” he said.
When it comes to audits, KPMG has been using base levels of AI for years in what it calls “data preparation,” he said. The technology is able to read and extract unstructured data, such as taking an invoice’s number, amount and data, from an image of a paper invoice, he said.
Now, with MindBridge’s technology the use of AI will be kicked up another notch to analyze rather than just prep it. For example, KPMG can harness the next level of unsupervised machine learning to analyze anomalies and unusual behavior in a company’s general ledger in a way that wasn’t possible in the past.
“For example, unexpectedly revenue transactions are debiting or crediting an account that they usually don’t, so the machine would identify the unusual transactional behavior compared to all the other revenue transactions that are following a more normal pattern,” he said.
As such, the auditing process shifts from a system that identifies issues through random sampling of transactions by using technology to hone in on transactions that may be more likely to warrant attention. That benefits the CFOs of the companies being audited too, because it saves their finance team’s time by avoiding the need to discuss routine issues.
“It allows us to focus our attention on the transactions, the areas of the audit where there is increased risk through unexpected unusual behavior and also we focus our client’s time into a much more valuable area,” Stöckle said, adding that the data driven approach transitions some of the rote audit workload from the business finance and accounting department to IT.
Still, in spite of all the advancements, Stöckle doesn’t foresee a time when audits will be performed without accountants.
“Nobody at KPMG sees that happening,” Stöckle said, reiteriating the importance of always maintaining a robust human-machine interaction. “Yes, the machine takes work away but that...allows humans to focus on those areas that are really important, potentially of higher risk.”