Are you using AI in your audit process? How so?
Sort by:
the webcast that Gartner did on GenAI in the audit process last week was really great - I think there is a replay or follow up in May, I'd highly recommend. There are real-life client examples from BIMBO and others
Can you share the details on that?
Yes we#ve developed a full AI audit assisitant that guide auditors through the audit workflow process making recommedations on what the scope of an audit should be, examining fieldwork documents to identify deficiencies and writing audit reports.
Thank you again for sharing your AI tool in the webcast. I'm looking at potential use case. As we are a team of 5, and based on your experience, what would be some prerequisite for a successful implementation and initial use case?
One of the key considerations is data security. We could only develop the system we have at E.ON as we have our own ringfenced Chat GPT system which can only be accessed by the Audit team. Therefore I'd suggest before thinking too much about use cases look at what foundations you need to put in place first to allow you to utilise the system.
From my conversations with AI Engineers particularly in insurance industry - Generative AI can be effectively used to identify inconsistencies in the set of documents, which can occur due to error or fraud. That can be done these days with today's technology.
We are using AI to increase the efficiency of auditing medical claims for accuracy. It is used to probabilistically filter out claims that are least likely to have an actionable error. It allows us to review more claims than we have resources available. Basically, it leverages machine learning to correlate a large database of historic claims and past errors against a large set of claim characteristics. Claims are scored for the probability of having an error and only the highest scored claims are reviewed.
We are currently using SAP Signavio to move our audit process from periodic sampling to Continuous Control Monitoring (CCM).
While we do use it to test workflow designs, the real value for us has been in Conformance Checking. We feed 100% of our transactional data into the system, and the AI compares the actual execution against our designed workflow.
Instead of just testing if the 'happy path' works, the AI highlights the 'unhappy paths', specifically flagging retroactive POs, approval bypasses, or cycle-time anomalies that a human auditor might miss in a sample set. It’s helped us shift from reactive auditing to near real-time monitoring.