CAQ 审计合伙人脉搏调查:AI 采用超前于治理,审计面临新风险
With AI adoption outpacing governance, audit faces new risks
CAQ 最新审计合伙人脉搏调查显示,57% 的审计合伙人将 AI 治理列为客户首要挑战,十年来首次超越网络安全。51% 将数据分析与 AI 列为今年首要任务,但近 90% 企业的 AI 治理成熟度仍处于「发展中」或「早期」阶段。66% 预期客户将自愿增加 AI 相关披露,重点应集中在证据评估与职业怀疑两个高风险环节。
For the first time in years, cybersecurity isn't the top concern for audit committees. The CAQ's recently published Audit Partner Pulse Survey found that 57% of audit partners now rank AI governance as their clients' single biggest challenge, ahead of cybersecurity and finance transformation.
Processing Content
I have spent seven years building Auditable AI for this exact boardroom conversation: technology designed to hold up under an auditor's scrutiny, not just accelerate the work. Cybersecurity held that top spot for a decade. Then AI took it. That alone should change how boards view the rest of 2026.
Adoption is outpacing oversight
Fifty-one percent of audit partners call data analytics and AI a top priority for their clients this year. Yet nearly 90% of companies sit at "developing" or "early stage" AI governance maturity, highlighting the challenge firms face in measuring how they use a technology the majority of our industry is leaning into.
The speed of adoption, while governance catches up, has peer reviewers and PCAOB inspectors watching closely, and the gap between the two won't change on its own, even if adoption slows. If anything, it's more likely to widen as more firms move faster than oversight can keep pace.
What CFOs should actually be disclosing
Sixty-six percent of audit partners expect their clients to voluntarily increase AI-related disclosures this year. These disclosures should do more than confirm a company uses AI. They should show where AI actually touches the audit, and what stands behind it: the source document, the system record, the standard it was tested against. That traceability is what we mean by Auditable AI. Two areas where AI adoption introduces the most risk, evidence evaluation and professional skepticism, are exactly where that disclosure should concentrate.
Real disclosure has to map to where the actual risk concentrates. If a company's AI disclosure doesn't explain how evidence is verified or where human judgment still sits in the process, it isn't telling investors anything they can actually use.
Two variables moving at once
Thirty-nine percent of clients are now seeking guidance on expected regulatory change, up from 23% in fall 2025. Companies are adjusting to AI adoption and a shifting SEC regulatory posture inside the same budget cycle. Having to absorb two major shifts is a tricky dance for any organization, especially those operating in a highly regulated industry like ours.
I've written before that AI adoption requires a different operating mindset than a typical software rollout. It delivers value through real-world use, continuous feedback, and governance. This is what we mean by Auditable AI. The finance teams best positioned to manage this movement are those that started with a small number of high-impact use cases, demanded explainable and auditable outputs, and scaled deliberately as confidence built, keeping human judgment central to the process the whole way through. That discipline is what makes it possible to absorb two major shifts, AI adoption and regulatory change, without one derailing the other.
The workforce side of the same problem
When AI takes over manual work, data entry, reformatting and reconciliations, the people who remain need different skills. Judgment, analysis and the ability to direct and verify AI output rather than produce it by hand will be the skills that the next generation of professionals must master. That explains why 54% of companies list targeted upskilling as a top human capital priority this year, while 58% are pursuing headcount reductions.
This shift means more meaningful work for young professionals, giving them more opportunities to add value than tick and tie their way through the early years. This requires firms to change their training and development programs and invest in deliberately building that judgment rather than assuming it develops on its own.
Closing the gap requires domain expertise, not just policy
Governance too often gets waved off as a backseat driver, told to sit back and let adoption do its thing. But governance isn't a backseat driver. It's the one keeping the car on the road.
Most firms are trying to close the adoption-governance gap with governance frameworks and disclosure checklists, but a policy is only as good as the people applying it. AI governance in audit is much different than AI governance in marketing or customer service. It requires people who understand what AS 1105 actually demands of evidence and where a model's confident output can quietly diverge from what a standard requires. A committee without that expertise in the room will write policies that look complete but miss the failure modes that matter.
As the survey points out, a firm can have a governance framework on paper and still be at an "early stage" in practice because people without the domain knowledge built it and didn't know what to test for. Solving this challenge requires putting practitioners in the room. Ffirms that treat it as a staffing and expertise question will move faster than those that treat it as a compliance exercise. The technology only works when the people who understand the standards stay in the loop.
What this means through the end of the year
The CAQ's latest data puts AI governance in the spotlight for audit firms. A few things worth putting on the agenda now:
- Track governance maturity as an ongoing metric, revisited every quarter, not a project you finish once.
- Map AI disclosure to where the real risk sits: evidence verification and human judgment.
- Build regulatory uncertainty into the budget as a recurring line item.
- Pair every AI investment with a deliberate upskilling plan.
- Staff governance committees with practitioners who know the standards.
- Consider governance from the start of any purchasing conversations. Ask any AI partner to show, not just tell, how its outputs trace back to source evidence.
Adoption has the wheel, and boards have noticed. The firms that win this cycle won't be the fastest drivers. They'll be the ones who kept governance out of the back seat and let it navigate.

President and Founder
Isaac Heller is President and Founder of Trullion, the leading AI-powered accounting and audit platform. Trullion works with Big 4 professional... Read full bio
来源:Accounting Today(美国会计行业媒体) · accountingtoday.com