A June 2026 survey of more than 1,000 audit professionals found that 88% now say AI risks undermining professional judgment. That’s not a fringe worry — that’s the profession’s own supermajority saying it out loud. Former PCAOB Acting Chair George Botic put it even more bluntly: AI, in his words, may compromise reasoning and judgment.
So why is this happening — and more importantly, what do you actually do about it?
A Pattern That’s Older Than AI
There’s a well-documented phenomenon in human factors research called automation complacency. It shows up any time a skilled professional starts relying on a system that’s usually right. The core finding, replicated across decades of aviation and radiology research: the more reliable a system becomes, the less actively humans monitor it — even when they’re technically still “supervising.”
Commercial airline pilots on long-haul flights are the classic example. Autopilot handles the plane for the vast majority of the flight, and it’s extremely reliable. But researchers have found something counterintuitive: pilots who spend years relying on autopilot show measurably slower reaction times when something rare actually goes wrong. It’s called the startle effect. The skill of actively monitoring — constantly cross-checking instruments, staying mentally ahead of the aircraft — atrophies when it isn’t exercised, because the brain reasonably stops spending effort checking something that’s almost never wrong. The pilot hasn’t gotten worse at flying. They’ve gotten worse at catching the rare moment the system actually needs them.
Radiology shows the identical pattern. When AI-assisted screening tools flag likely abnormalities, radiologists using those tools show real, measured drops in their own independent detection rate for the cases the AI happens to miss — not because the doctor lost skill, but because attention naturally reallocates toward confirming what the system already flagged, rather than independently forming a diagnosis first and comparing.
That’s the mechanism. It’s not a character flaw, and it’s not a training gap you eliminate with firmer language in an engagement letter. It’s a predictable consequence of trusting a system that’s usually right — and it will happen to good auditors just as surely as it happens to good pilots, unless something specific counteracts it.
The Profession — and the Classroom — Are Already Naming It
Two things are converging right now. First, that 88% figure means the people doing the work already sense something is off. Second, accreditation standards are responding. AACSB — the Association to Advance Collegiate Schools of Business, one of the “big three” global business school accreditations — created a new standalone accounting standard called Standard A5: Digital Agility, effective July 1, 2026. The language is direct: learners must be able to “evaluate the appropriateness of technologies for specific accounting tasks, interpret technology-enabled outputs with accuracy and skepticism, and communicate findings clearly and ethically.”
That’s the global accreditor of business education making skepticism toward AI output a formally assessed, required competency — not a suggestion in a white paper. And the detail that matters most for those of us already in practice: research on how skepticism is taught shows that a questioning mind and the drive to verify measurably improve through deliberately designed instruction. Skepticism isn’t a fixed trait — it’s trainable. Which means it’s also something that decays without deliberate practice, exactly like a pilot’s manual flying skill on autopilot.
The Fix: Form Your Expectation First
The antidote to automation complacency isn’t distrusting the system more — it’s preserving your independent anchor before you ever look at the system’s output. In aviation, this is called staying ahead of the automation: a pilot doing it well has already formed a mental model of what the aircraft should be doing next, before checking whether it actually is.
Translate that into your workflow. Before you open the AI-drafted analysis, the AI-flagged anomaly, the AI-suggested account treatment — pause and form your own expectation of what the answer should roughly look like and why. Then compare. If they match, that’s real confirmation, because you generated an independent judgment first. If they don’t match, that’s the moment professional skepticism actually kicks in — because now you have something to measure the AI’s answer against.
The failure mode is opening the AI’s output first and letting it become your starting point. Once that happens, everything afterward is confirmation of what you already saw, not independent verification — the exact gap regulators and accreditors are now pointing at.
Key Takeaways
88% of surveyed audit professionals already believe AI risks undermining professional judgment.
Automation complacency is decades-old, well-documented human factors research from aviation and radiology.
AACSB’s new Standard A5, Digital Agility, formally requires accuracy and skepticism when interpreting AI output, effective July 1, 2026.
The fix: form your independent expectation before you see the AI’s output, not after.
Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: https://everydaycpe.com/courses/auditing/automation-complacency-and-professional-skepticism/
Leave a Reply