MindBridge urges clearer PCAOB guidance on AI audits

MindBridge has submitted formal comments to the Public Company Accounting Oversight Board on artificial intelligence in audit oversight, urging clearer guidance on technology-assisted audit work.

The filing argues that current oversight frameworks do not fully address how audit firms should document, assess and defend AI-assisted procedures during inspections, especially when they analyse full populations of transactions rather than selected samples.

MindBridge, which develops AI tools for financial analysis in audit work, argues that the PCAOB should make AI-augmented procedures an explicit strategic priority. Its response focuses on how the regulator's inspection, standard-setting and reporting programmes should adapt as automated analysis becomes more common in audit workflows.

According to MindBridge, the audit profession still operates under standards built around sampling and manual review, while newer systems can assess entire transaction populations and assign risk scores across them. It argues that this shift has created uncertainty for firms that want to rely on technology-assisted work but are unsure how inspectors will judge the resulting evidence.

Inspection gaps

A central issue in the submission is the lack of detailed guidance following the PCAOB's amendments on technology-assisted analysis. MindBridge says auditors and software providers still face unresolved questions about how to set investigation thresholds when every transaction receives a risk score, and how to treat lower-risk items when deciding whether sufficient audit evidence has been obtained.

It argues that this uncertainty can lead to inconsistent inspection outcomes, with firms carrying out similar work receiving different treatment depending on the inspection team reviewing an engagement.

MindBridge adds that this encourages duplication rather than better audit work. Firms may run AI-assisted tests alongside traditional manual procedures because they do not know whether technology-based work alone will satisfy inspectors.

"Technology is not a supplement to good auditing. It is increasingly the means by which good auditing is achieved," said Wenzel Reyes, Head of AI Governance at MindBridge.

MindBridge also calls on the PCAOB to publish aggregated, anonymised information on the types of technology-assisted procedures inspectors encounter and the deficiencies they identify. It says such reporting would help firms understand where implementation is falling short and provide a clearer picture of how the amended standards operate in practice.

Standards review

Beyond immediate guidance, the submission presses for a broader review of auditing standards that assume procedures are carried out and supervised by people rather than software. MindBridge argues that the regulator should examine supervision rules, documentation requirements and quality management standards to determine how they apply when AI systems perform parts of the work.

It also calls for coordination with international standard setters, including the International Auditing and Assurance Standards Board and the International Ethics Standards Board for Accountants. Cross-border firms, it says, should not face conflicting requirements for how technology-assisted procedures are documented or assessed.

Another part of the response addresses the PCAOB's own use of technology. MindBridge argues that effective oversight in an AI-driven environment will require more than software procurement and will depend on training inspectors to assess whether judgment has been applied properly to machine-generated findings.

"An inspection model that treats a well-documented, AI-assisted full-population test identically to a manually selected sample of several items in a world of big data is not evaluating audit quality - it is evaluating process familiarity," Reyes said.

UK rollout

The submission comes as MindBridge expands in the UK audit market through a partnership with BDO UK. The accounting firm is rolling out MindBridge's analytics platform across its practice after a trial, aiming to increase the use of data-led audit methods ahead of year-end reporting periods.

The arrangement reflects a broader shift among audit firms towards testing entire datasets rather than relying solely on smaller samples. Supporters argue that this approach can improve the identification of unusual transactions and financial anomalies, though regulators and firms still need common standards for interpreting and documenting the results.

MindBridge frames its intervention as part of a wider debate over how regulators should respond as AI becomes embedded in core assurance work rather than remaining an optional add-on. It argues that strategic planning for the next phase of audit oversight should reflect that reality if investor protection and audit quality are to be maintained.

In closing, MindBridge says it wants to contribute practical experience from audit engagements across several countries as the regulator refines its approach to AI. "A profession that adopts technology without the standards, oversight frameworks, and inspector competencies to evaluate it serves no one's interests, least of all investors," said Reyes.