How Is AI Transforming Audit and Assurance Services?

AI is transforming audit and assurance in three ways: automating evidence-gathering tasks such as document review and transaction testing, enabling full-population analysis instead of sampling, and reshaping client expectations, since 72% of companies already use or pilot AI in their own financial reporting. The change is real but uneven, and governance is lagging adoption.

That short version hides a lot of nuance. Adoption numbers are rising fast, yet most firms are still automating the edges of the audit rather than its core judgments. What follows is what the verified data says, where AI is actually being used across the audit cycle, and what managing partners should do while the technology, the standards and the economics are all moving at once.

Key Takeaways

• Generative AI use in tax and accounting firms nearly tripled in a year, from 8% in 2024 to 21% in 2025, and 71% of professionals now believe GenAI should apply to their daily work (Thomson Reuters, 2025) [1].
• Clients are ahead of many auditors. 72% of large companies are piloting or using AI in financial reporting, expected to reach near-universal levels within three years (KPMG, 2024) [2].
• 82% of companies believe their auditors are ahead of or level with them in adopting AI for financial analysis, which sets a high expectation bar (KPMG, 2024) [2].
• The strongest near-term use cases are evidence extraction, full-population journal entry testing, risk assessment analytics and drafting support, not automated judgment.
• Governance is the weak point. Many firms adopted tools before policies, and regulators including the PCAOB and FRC have signalled close interest in how AI-assisted audits are supervised and documented.

The Adoption Picture, in Verified Numbers

It helps to separate three groups: professional services firms, their clients, and the audit function specifically.

Across professional services broadly, Thomson Reuters’ 2025 Generative AI in Professional Services report found 22% of organizations actively using GenAI, up from 12% a year earlier. Within tax and accounting, adoption nearly tripled from 8% to 21%, and the share of professionals who believe GenAI should apply to their daily work rose from 52% to 71% (Thomson Reuters, 2025) [1].

Clients are moving at least as fast. KPMG’s global study of 1,800 companies across ten major markets found 72% already piloting or using AI in financial reporting, with adoption expected to reach 99% within three years. 82% of those companies believe their auditors are ahead of or level with them in applying AI to financial analysis (KPMG, 2024) [2]. That belief is a commercial asset for the profession, and a fragile one. An auditor who arrives with manual sampling and PDF tick-marks in front of a finance team running AI-assisted close processes is spending reputational capital every visit.

In our own research at Phronesis Partners for professional services clients, we see the same asymmetry from the buyer side. When we interview CFOs and audit committee members, expectations of AI-enabled audit are usually formed by what their internal teams already do rather than by what audit standards currently permit. Managing that expectation gap is becoming part of the engagement relationship itself.

Where AI Is Actually Being Used Across the Audit Cycle

The phrase “AI audit” risks sounding like a single product, so it is worth being concrete about where the technology sits today.

Planning and risk assessment

• Analytics over full general ledgers to flag unusual patterns, ratios and trends before fieldwork starts.
• External data scanning across news, filings and sector data to inform the team’s understanding of the entity and its risks.
• GenAI summarization of prior-year files, board minutes and contracts to speed up team onboarding.

Evidence gathering and testing

• Full-population journal entry testing with anomaly scoring, replacing or augmenting sample-based approaches.
• Intelligent document extraction, pulling terms from leases, loan agreements and invoices for agreement to recorded amounts.
• Confirmation and reconciliation matching at scale.

Completion and reporting

• Drafting support for memos, summaries of misstatements and internal communications, always subject to reviewer sign-off.
• Consistency checking across the financial statements and disclosures.

Two honest caveats belong here. Most firms outside the largest networks are still in the first category, planning analytics and drafting support, rather than deep evidence automation, because the latter needs standardized data pipelines few clients can feed. And none of the credible deployments remove human judgment on materiality, estimates or going concern. The technology narrows where judgment is needed. It does not supply it.

Old Model and AI-Enabled Model

The pyramid row deserves attention from anyone who owns a staffing model. If routine testing shrinks, the traditional leverage model of many juniors and few partners changes shape, and so does the way juniors learn. Several practice leaders we have interviewed worry less about the technology than about where the next generation of skeptical seniors comes from if the grunt work that trained them disappears. There is no settled answer yet, though firms experimenting with rotational data-analytics roles and simulation-based training are at least asking the right question.

The Risks Regulators and Audit Committees Are Watching

Adoption is outpacing governance, and that gap is where the profession’s risk sits. Thomson Reuters found that many firms lack formal GenAI policies even as usage climbs (Thomson Reuters, 2025) [1]. The recurring themes:

• Hallucination and over-reliance. GenAI outputs read confidently whether or not they are right. Audit methodology has to treat them as unverified input, with documented human review.
• Data confidentiality. Client financial data cannot flow into consumer-grade tools. Firms need contained environments and clear usage policies.
• Explainability and documentation. Inspectors will ask how a conclusion was reached. “The model flagged it” is not audit evidence. The workpaper trail must show what was done with the flag.
• Bias and completeness in training data. Anomaly models tuned on one sector’s patterns can miss another’s.
• Auditing the client’s AI. As clients automate reporting, auditors must assess controls over models they did not build, a skills demand arriving faster than the training catalogue.

Regulators are engaged rather than obstructive. The PCAOB and the FRC have both published staff commentary on AI use in audit and are gathering information on firm practices, and the improving inspection results discussed in our companion piece on audit market trends suggest technology and quality can rise together. Even so, the first enforcement case involving an unsupervised AI output will be studied line by line across the profession.

What Practice Leaders Should Do Now

The strategic question has shifted from whether to adopt AI to how fast, in what order, and with what evidence about clients and competitors. A practical sequence we see working:

1. Baseline your clients’ expectations. Survey and interview buyers directly. Their AI maturity, not your roadmap, determines when manual delivery starts to look dated. Phronesis Partners runs this kind of demand-side research using proprietary B2B panels that reach CFOs, controllers and audit committee members.
2. Benchmark competitors’ actual capability. Press releases overstate. Recruitment adverts, tool partnerships and client-side interviews reveal what rivals really deploy, and competitor benchmarking separates marketing from method.
3. Pick two or three use cases with measurable hours saved. Document extraction and journal-entry analytics usually pay back first. Track realized hours, not vendor promises.
4. Write the governance before scaling. Usage policy, approved tools, review requirements, documentation standards. Cheaper now than after an inspection finding.
5. Rebuild the training pathway. Decide deliberately how juniors will develop skepticism when routine testing is automated.

The economics deserve one plainly commercial point. If AI compresses audit hours, firms that price purely on time will hand the entire efficiency gain to procurement. Firms that reprice around assurance value, coverage and speed will keep some of it. Which pricing arguments land with buyers is a researchable question, and a better use of a strategy budget than another internal debate.

Frequently Asked Questions

Will AI replace auditors?

No credible evidence points that way in the foreseeable future. AI automates evidence gathering, testing and drafting, while professional judgment on risk, estimates, materiality and going concern remains human and legally accountable. The likelier outcome is smaller, differently skilled teams, with 71% of tax and accounting professionals now expecting GenAI to be part of daily work (Thomson Reuters, 2025) [1].

How widely is AI actually used in audit today?

Generative AI use in tax and accounting firms roughly tripled from 8% in 2024 to 21% in 2025 (Thomson Reuters, 2025) [1], and on the client side 72% of large companies are piloting or using AI in financial reporting (KPMG, 2024) [2]. Most audit deployments today focus on analytics, document extraction and drafting support rather than automated judgment.

What do clients expect from AI-enabled audits?

KPMG’s 2024 global study found 82% of companies believe their auditors are ahead of or level with them in AI adoption for financial analysis. In practice, clients increasingly expect full-population testing, fewer repetitive document requests, faster turnaround and insight drawn from their data, and they notice when a firm cannot deliver them.

What are the biggest risks of using AI in audit?

Over-reliance on unverified outputs, client data confidentiality, inadequate documentation of how AI-assisted conclusions were reached, and gaps in firm-level governance. Regulators including the PCAOB and FRC are actively reviewing firm practices, so policies and review controls should be in place before tools are scaled.

Phronesis Partners is a global market research and insights firm. We help audit, accounting and consulting firms track technology adoption, benchmark competitors and understand buyer expectations through full-service qualitative and quantitative research and proprietary B2B panels.

References

1. Thomson Reuters, 2025 Generative AI in Professional Services Report, 2025, https://www.thomsonreuters.com/en/reports/2025-generative-ai-in-professional-services-report
2. KPMG, AI Transforming Financial Reporting Globally, 2024, https://kpmg.com/xx/en/media/press-releases/2024/05/ai-transforming-financial-reporting-globally-with-near-universal-adoption-expected-in-the-next-three-years.html