How skepticism of AI could raise audit transparency and reproducibility

# The skepticism problem and the unexpected benefit

# What auditors are asking about AI

Conversations about AI in audit repeatedly circle back to traceability questions: Can the process be reperformed? Is every conclusion traceable to source data? Can a reviewer determine exactly how a result was generated? Is a complete audit trail retained? These are valid questions, and their repeated emergence is creating pressure to improve how audit procedures are documented — both for AI-assisted and traditional human processes.

# Where current audit practice falls short

Experienced auditors frequently rely on accumulated experience, intuition and judgment. Final workpapers often show selected items and conclusions but not the full chain of reasoning. That has been accepted practice because professional judgment is inherently hard to fully reconstruct. AI is forcing firms to ask whether more of that invisible reasoning can be captured, structured and reproduced.

# Practical changes driven by AI skepticism

• Retain source populations and selection criteria for sampling driven or assisted by AI so selections can be independently reperformed.
• Document inputs, assumptions and outputs for AI-assisted analyses to give reviewers a clearer picture of how evidence was evaluated.
• Build audit trails that connect evidence directly to outcomes rather than relying on implicit professional experience.

These measures are not AI-specific. They point to improving documentation and reproducibility across audit procedures, which can reduce variability between engagement teams and make reviews more effective.

# How transparency improves reviewer oversight

When AI-assisted work requires explicit retention of data, methodology and results, reviewers gain heightened visibility into how conclusions were reached. That helps in two ways: it lets reviewers reperform steps when necessary, and it makes it easier to evaluate whether professional judgment was applied appropriately. The increased granularity narrows the gap between a documented conclusion and the underlying reasoning that led there.

# From skepticism to constructive adoption

Skepticism can be a barrier to adoption, but it can also drive improvement. The line of questioning aimed at AI—about reproducibility, traceability and audit trails—encourages firms to treat those attributes as non-negotiable. By demanding these features of AI-assisted workflows, firms end up designing processes that raise documentation standards enterprise-wide.

# Bottom line