In this ever in increasing AI world nothing is potentially as it seems.  For example invoices or receipts for expenses can now easily be artificially created to look like the real thing. To me this is changing the risk profile of audits and evidence we could previously rely on to back up process compliance may no longer be authentic.   I'd be interested to know how audit teams are responding to this and are you changing audit processes to reflect that collected evidence may not always be as it seems?

I think you would really need to evaluate if the risk of the potential fraudulent source document is worth changing your processes. Expense reporting is an area that analytics use is helpful and we would utilize this before creating a new audit test. 

Publish the latest list of all different service providers with complete details (both for approved and not approved ones) published and your models to learn from that. This can help them to identify fraud cases. Also ensure in more frequent sample audit checks to validate the quality of assessment done by your AI agents. 

Implementing our AI tool to review claims we have AI check the validity of the service, was it a real doctor, is it an existing address, it is an existing pharmacy, etc. This can be done a lot faster thru AI than manually and likely provides better protection against fraud.