When I worked as an auditor, I wasn’t sitting around wishing I had artificial intelligence to help me test evidence. Frankly, I knew nothing about AI. What I did know a LOT about was the pain of trying to get through mountains of tedious administrative work. If you’re also in the auditing world, you know exactly what I’m talking about.
There were countless emails to send to clients, spreadsheets to reconcile, documents to copy and update, and requests to track across multiple tools that didn’t communicate with one another. We repeated work from one audit to the next, sometimes copying controls from a prior-year report, updating them, sending them to a client, and then reconciling the returned information against what had changed. Although my auditing days are behind me, I still occasionally still have a nightmare or two about this part of the process during the busy season.
Even when the actual audit work was complete, there was still a tremendous amount of manual cleanup. Spreadsheets crashed. Information had to be re-entered. A change in one place could mean updating several other documents.
Looking back at the broken processes, I don’t think the answer is to shove AI into auditing for the sake of saying we’re “AI-powered.” I think we should be asking a much more important question: “What work should technology take off an auditor’s plate, and what responsibility needs to stay with the auditor?”
A(I) is for Admin
Start with the work that is repetitive, structured, and administrative. That’s where AI and automation can make an immediate difference in the audit process. AI should help organize evidence, compare documents, identify changes, flag missing information, prepare follow-up requests, and help generate documentation from information that already exists in the audit workflow (which is exactly what we’re building at Yak Tech).
These tasks that consume time, create opportunities for error, and often pull auditors away from the work that requires their expertise.
Consider a control that maps to multiple requirements or appears in more than one report. An auditor may need to test it once, then ensure that the relevant information is reflected consistently across the appropriate documentation. If the original control changes, those copies may need to be updated, too.
That is exactly the kind of process that technology should and can help manage. It is structured work with clear relationships between the underlying information and the documents that use it.
The same is true of reconciliations and first-level reviews. AI can identify discrepancies, compare information, and surface items that deserve attention. It can act as an additional set of eyes (provided the auditor can see what it reviewed and verify the result).
H(uman) is for Hold Up
However, I want to caution that performing a task is not the same as owning the outcome. And that is where I find I fundamentally disagree with many of the technology companies entering our space about AI’s place in the audit.
An AI system may be able to analyze a document, identify a possible exception, or recommend that evidence meets a requirement. That does not mean it should independently conclude that a control passed, determine that evidence is sufficient, dismiss an exception, or issue an audit conclusion without meaningful auditor review.
Auditing is a chain of evidence, judgment, and professional accountability. The auditor needs to understand what was tested, which evidence was used, whether it covered the right period and population, what assumptions were made, and how exceptions were handled.
A final approval button does not automatically make a process meaningfully human-led. The auditor must be able to evaluate the work, challenge the output when necessary, and remain accountable for the conclusion.
That is especially important because audit work is not always standardized in the way technology companies might expect. Different firms have developed their own methodologies and approaches to evaluating whether evidence sufficiently addresses a risk or control objective (and that is why I’m a proponent of BYOW). Technology should support those methodologies rather than force every firm into one prescribed process.
The Audit Trail is Not Optional
In an audit, you have to be able to defend the work that was completed. Now is not the time to hand it all over to the AI agent. If an AI tool produces a recommendation, the auditor should be able to understand what information informed it, review the underlying evidence, and evaluate how the result was reached.
A system that simply accepts a file and produces a polished report may look impressive in a demonstration. But an auditor needs more than a polished output. They need visibility into the evidence, the process, and the exceptions.
Trust depends on integrity. If information goes into a system, the auditor needs confidence that it remains accurate and traceable throughout the workflow. Any unexplained change to data, controls, or testing notes should raise questions.
Use the Tool, but Keep the Humanity
The goal should be a more human audit process, not a less human one. I want to be clear that AI and automation are not substitutes for professional skepticism in our industry.
I truly believe AI and automation can help auditors spend less time chasing evidence, organizing files, copying information between systems, and repeating administrative work, and that’s why I built AI capabilities as a fundamental portion of the Yak infrastructure. This way, we create more room to investigate risks, understand exceptions, communicate with clients, and apply professional judgment.
Proper infrastructure can also help address a capacity problem that hiring alone cannot solve. Adding people to inefficient workflows increases cost without necessarily increasing capacity proportionally. And developing an auditor who can independently recognize risk, evaluate unusual situations, and make sound judgments takes years of experience.
But I don’t believe the answer is to remove the human relationship from the process. Unlike some of the companies building audit AI, I don’t believe we need to replace auditors with AI. Auditors are not simply there to catch clients doing something wrong. Their role is to evaluate evidence objectively, communicate findings, and help organizations understand where processes need to improve.
The best use of AI in audit is not to make the auditor disappear. It is to remove the mechanical work that gets in the way of the auditor doing the job well.
ABOUT THE AUTHOR:
Shannon Smith is the founder and CEO of Yak Tech, an automation and artificial intelligence platform for audit and accounting firms, and a former SOC auditor.
Photo credit: Freepik
Sign in to get access to this free resource, and all of our whitepapers and reports.
Download this content today!
Register Now Already registered? Click here to Log In
