Artificial intelligence (AI) can enhance audit quality, but it can also produce incorrect results. That dichotomy introduces fresh challenges for audit professionals seeking to leverage AI on their engagements while continuing to adhere to professional standards.
“AI is changing the way we do our work, but it’s not changing our professional requirements with respect to professional skepticism, use of judgment, appropriate supervision, and review,” said Jennifer Burns, CPA, the AICPA’s chief auditor.
Although today’s models offer great potential for the audit profession, they also tempt human operators to abdicate their own judgment and accept AI’s conclusions as truths. That poses a huge risk because AI delivers outputs ranging from surprisingly smart to fatally flawed with the same confidence.
“As auditors, the capital markets rely on us to provide them with competent, independent assurance. If overreliance on AI results in auditors failing to detect material misstatements, for example, that could put the value of the audit in question and the trust the profession has earned over the past 100-plus years at risk,” said Carl Mayes, CPA, AICPA vice president—Ethics and Firm Quality.
The ultimate goal is not just to put a human in the loop, Mayes said. Instead, he added, auditors should aim to put the human in the lead, with clear responsibility and accountability for their engagements even as AI’s capabilities expand.
To provide more detailed guidance on how auditors can use AI tools without running afoul of professional standards and ethics, the AICPA Ethics Division in March launched an Ethics Staff Insights article (see “AI Through an Ethics Lens,”). Also, in October, the staff of the AICPA’s Assurance & Advisory Innovation and Audit & Attest Standards teams and the Assurance Services Executive Committee AI Working Group were scheduled to issue a practice aid titled Artificial Intelligence (AI) Considerations for Financial Statement Auditors.
UNDERSTANDING AI’S POTENTIAL FOR THE AUDIT
AI’s fast-changing nature has left many auditors with questions about how and whether they may use various AI tools.
Answers to this technological challenge are embedded in the AICPA Professional Standards, including the AICPA Code of Professional Conduct. For auditors leveraging AI, adhering to those requirements means understanding AI’s fundamentals, testing and challenging its outputs, governing its usage, and appropriately training staff.
Auditor supervision of AI tools must match the rigor applied to human staff, according to Mayes. “Imagine an AI tool is a very bright first-year associate,” he said. “They may do great work, but ultimately, they require diligent supervision and review.” And that’s where quality management responsibilities under the AICPA standards come into play, Burns added.
Various forms of AI can be applied across the audit process, from streamlining repetitive tasks to generating evidence. According to the draft AI practice aid, AI may be used in audits for:
- Analyzing financial, operational, industry, and prior–period audit information to identify areas, accounts, assertions, or transactions that may present elevated risk.
- Extracting text from scanned invoices, receipts, and paper documents so the information can be used in audit procedures.
- Cleaning, formatting, and standardizing financial data from different client systems into a common format for auditor review.
- Analyzing financial and operational data across periods, business units, products, customers, or locations to identify relationships and trends that are inconsistent with expectations.
- Analyzing contracts, agreements, minutes, and other documents to identify key terms, risks, obligations, or matters relevant to the audit.
- Comparing information across multiple documents to identify inconsistencies, omissions, or matters requiring auditor follow–up.
- Comparing draft disclosures to applicable disclosure requirements, accounting guidance, or a checklist to flag missing, incomplete, or inconsistent items.
Each of those uses requires familiarity with the technology, careful planning and consideration, and vigilant testing and skepticism. That means that auditors should evaluate their new tools, using quantitative and qualitative checks to, for example, confirm that AI-generated transcripts have a high accuracy rate or that an AI-agent-powered workflow delivers uncorrupted and reliable data.
“With the pace of change only accelerating, CPAs have to change how they do work,” said Jeremy Dillard, CPA, CGMA, technical standards partner at SingerLewak LLP and a member of the AICPA Auditing Standards Board.
Larger firms are constantly experimenting, he said, including by introducing new tools for some audits and seeing how they work.
“It’s probably at a point where, once a quarter, we’re looking at new audit technologies and rolling them out,” Dillard said.
WHAT THE STANDARDS SAY
Auditing and quality management standards recognize the use of “technological resources” in an audit, but they don’t tell a practitioner whether to use AI. Instead, they provide a principles-based approach to managing the use of technology and considering its outputs.
“Technology is a tool,” Burns said. “As CPAs, we can’t abdicate our responsibility to AI. Instead, we need to leverage AI to help us achieve our objectives and meet our professional requirements.”
That’s a message a majority of the profession supports worldwide. According to a 2026 IDC survey of about 1,000 respondents working in audit and assurance in six countries, 53% agreed or strongly agreed that AI tools can enhance audit quality, and 64% said they believe auditors should always validate the AI results they rely on to reach their professional conclusions (see “Using Technology to Boost Audit Quality,” JofA, Jan. 1, 2024).
According to AU-C section 220, Quality Management for an Engagement Conducted in Accordance With Generally Accepted Auditing Standards, the “engagement partner should determine that sufficient and appropriate resources to perform the engagement are assigned or made available to the engagement team in a timely manner, taking into account the nature and circumstances of the audit engagement.” This would include the use of technological resources.
That and other requirements may affect AI usage identified by the AI practice aid.
A BALANCED APPROACH TO AI
Firms should be wary of the marketing claims made by some AI vendors, which may encourage auditors to rush through engagements.
Mayes cited reported instances of individual partners performing thousands of system and organization controls (SOC) engagements per year using AI-backed platforms, raising questions about whether they met applicable supervision and review requirements (see “Promises of ‘Fast and Easy’ Threaten SOC Credibility,” JofA, Feb. 1, 2026).
“It really comes down to overreliance on vendors and those vendors’ technology, and subordination of professional judgments to those vendors and tools,” Mayes said.
Firms seeking to leverage AI in their engagements will want to establish effective governance processes, which many businesses are embracing amid AI transformations.
“It’s true for any tool, for any technology that we’re putting in place: You want to have the appropriate controls,” Burns said.
Effective AI governance includes:
- Processes and responsibilities for deploying AI.
- Policies and training for its use.
- Ongoing evaluation and monitoring of performance.
- Further controls related to access, audit trails, change management, and version control.
The proper use of AI begins with engagement partners, who should ensure they have a broad familiarity with the technology and its potential applications. “Make sure you have a broad understanding of the AI tool itself, the task performed, and how the tool accomplishes the task,” Burns said.
Burns cautioned that auditors shouldn’t always rush for the most advanced technology.
“Not every problem or issue needs AI,” she said. “If you have a well-defined set of rules and need consistent, repeatable results, then conventional software might do the job just as well.”
Firms also should be careful to examine offerings from third-party vendors, Dillard cautioned.
He recommended that firms ask themselves, “What are other firms using? What are the more popular tools in the industry, as a starting point? Has anyone issued a SOC report that gives me some confidence in the integrity of the platform itself? Has anyone examined this code or looked at the IT controls over it?”
Also, firms should apply skepticism and change-management controls to software tools generated in-house through AI-powered “vibe coding,” which means using AI to generate software code from prompts instead of writing it manually.
Once AI is in use, auditors can maintain their professional skepticism by evaluating the AI tool’s reliability. In doing so, the AI practice aid urges audit firms to evaluate the accuracy of the AI outputs, the completeness of the processing, the consistency of the results of repeated measurements over time, and the ability to obtain the same results under identical conditions (repeatability).
For example, in evaluating the accuracy of the outputs, a firm might test the AI’s output against a known dataset, ensuring that its calculations and methods are trustworthy. Auditors also should evaluate an AI model’s ability to plausibly explain, for example, why it flagged a particular transaction for review.
There’s also the question of data integrity. While new technology can enhance the audit by analyzing full populations, it also increases the risk of compounding errors if the underlying data is faulty. Firms also can test that data extraction and transformation processes are not deleting or corrupting records.
Finally, firms may or may not need to detail AI uses in audit documentation. In evaluating the auditor’s documentation, the core question is whether an experienced auditor having no previous connection with the audit would be able to understand the nature, timing, and extent of the audit procedures performed to comply with GAAS and applicable legal and regulatory requirements.
“AI doesn’t replace professional judgment; it has to be used with the same rigor and skepticism as any other tool,” Burns said. “I think more audit committees today are asking their auditors and their audit firms not only, ‘How are you using technology?’ but also ‘How are you managing the risks that come with it?’”
FIGHTING AUTOMATION BIAS
The draft AI practice aid warned of the dangerous human tendency to favor automated decision-making systems, even when there’s obvious evidence to the contrary.
Automation bias, which has been studied for decades, describes the risk that users will follow a faulty AI suggestion or fail to act independently if they aren’t prompted by the AI.
Responsible use of AI “means understanding how our minds work, and it means understanding the risk of relying on something that appears to be all-knowing,” Mayes said. In an audit, automation bias can result in a failure to critically assess audit evidence, the AI practice aid warns.
Practitioners can combat this bias through human validation of significant outputs of the AI tool, training and education, and by maintaining their overall posture of independence and skepticism.
It is also recommended that high-risk areas of the audit include at least one non-AI-based procedure as a “contrarian check.”
Interestingly enough, AI also can be useful in monitoring AI. One model can be asked to review the work of another, looking for mathematical errors, faulty logic, and more. But, ultimately, a human still oversees this increasingly complex loop.
“In such situations, the second AI tool does not reduce or take the place of the auditor’s responsibilities for supervision and review,” the draft practice aid stated. “While an auditor may employ one AI tool to monitor another AI tool, the auditor retains responsibility for supervision of the tools.”
A CHANGING PATH
AI comes with profound challenges and opportunities for auditors at all stages of their careers.
Younger auditors face the dilemma that automation is replacing basic work — the rote assignments where staff once learned the fundamentals (see “How Will Accountants Learn New Skills When AI Does the Work?” JofA, March 1, 2026). Without the opportunity to learn these systems themselves, staffers may struggle to overcome automation bias and challenge the output of an AI system.
“[For] employers, this is a top concern,” Mayes said. “Entry-level staff are relying on the output of the bot without challenging it — and that’s a concern for now, but also a concern for the future, because you need them to develop those critical thinking skills.”
In response, some of the largest firms, with initiatives such as KPMG TaxSIM, PwC Virtual Learning Experiences, Deloitte University, and EY Audit Academy, are using simulation-based training to familiarize new staff with basic processes, and all firms should redouble their efforts on training, Mayes advised.
“I think it’s training them on critical thinking very explicitly, training them on ethics and integrity and adopting a skeptical mindset,” he said. “And it’s also making sure they have the technical foundation so that they know when to call [it out].”
At the same time, Mayes reminded auditors to identify opportunities to consider how AI might augment, rather than subjugate, their judgment. An auditor equipped with more comprehensive and accurate data obtained through automation and AI will be better equipped to assess a company’s financial health and controls.
Burns compared the AI wave to previous technological advances. “Every major advancement in technology has made our work more efficient so we can focus on the most complex, judgment-intensive areas,” she said. “We’re not checking our brain at the door and handing our judgment over to a machine. The ultimate responsibility remains with us.”
About the author
Andrew Kenney is a freelance writer based in Colorado. To comment on this article or to suggest an idea for another article, contact Jeff Drew at Jeff.Drew@aicpa-cima.com.
LEARNING RESOURCES
Applying AI in A&A: Creating Effective Prompts
Learn how to use enhanced prompts to unlock more accurate results in your AI outputs and save even more time.
Nov. 19 and Dec. 17, 1–2pm ET
WEBCAST
NextGen Leadership: Using Critical Thinking to Guide AI-Driven Decisions
This session examines how critical thinking enhances the application of AI in modern finance, ensuring accuracy in month-end close, forecasting, automation, and compliance processes.
Dec. 8, 2–3pm ET, and Jan. 5, 11am–noon ET
WEBCAST
AICPA ENGAGE 27
The accounting and finance profession’s premier event brings together thousands of accounting and finance professionals from across the country. Attendees gain fresh insights, expand their networks, and accelerate their growth. Scan the QR code or go to aicpaengage.com.
June 7–10, ARIA Resort & Casino, Las Vegas
CONFERENCE
Ethics and Critical Thinking in an AI World
This is a highly interactive, real-world program that is designed for accounting and finance professionals who are integrating artificial intelligence into their work.
CPE SELF-STUDY
For more information or to make a purchase, go to aicpa-cima.com/cpe-learning or call 888-777-7077.
MEMBER RESOURCES
Articles
“Accounting Ethics in the Age of AI,” JofA, Aug. 5, 2026
“Drafting an AI Policy That Actually Works,” JofA, July 1, 2026
“5 Human Competencies CPAs Need in the AI Age,” JofA, May 1, 2026
“How AI Is Transforming the Audit — and What It Means for CPAs,” JofA, Feb. 1, 2026
“How Accountants Can Balance Technology and Critical Thinking,” JofA, April 1, 2025
