AI in audit is moving from "what is it?" to "which tool, how governed, and how do we prove it works?" Money is flowing into audit-specific platforms, new benchmarks show where models still fail on real audit work, India's professional body has made AI a priority, and searches are shifting toward tools and governance standards.
Last reviewed: 17 October 2026. Facts checked. ICAI's summit facts come from ICAI's own release. Fieldguide's funding comes from press coverage of its announcement. The two benchmark papers were read on arXiv. The search figures are our own pull from OpenSEO (India, Google), and are small numbers. Claims about the Big Four and about agentic platforms come from commentary, not company filings, and are labelled. We could not open the PCAOB's pages, so we draw no conclusion from them.
For announcements month by month, see our October 2026 AI in internal audit update. This post is about the direction of travel.
Signal 1: Funding is going to audit-specific AI
Fieldguide, an AI platform for audit and advisory firms, announced a $75 million Series C in February 2026, led by Goldman Sachs growth equity, at a reported $700 million valuation and about $125 million raised in total. One report says half of the top 100 US accounting firms use it; that comes from the company's own announcement, so treat it as a claim.
What it means: investors believe audit workflow is a big enough market for dedicated AI products. Expect more tools, and more sales pitches.
Signal 2: Large firms are reported to be moving from pilots to platforms
Commentary this year describes the Big Four moving from AI pilots to production platforms, with agent-style systems that run routine procedures. These accounts come from vendor and secondary blogs, not from the firms' filings, so we do not repeat specific numbers. Our own read of what this means for mid-sized Indian firms is in Big 4 AI in 2026.
Signal 3: Benchmarks now test audit work, and show the limits
Two new papers test models on audit-style tasks:
- FinAuditing uses real XBRL filings, with 1,102 annotated instances averaging over 33,000 tokens, and 13 language models. The authors report substantial gaps in retrieving the right concepts and in reasoning consistently across documents.
- AuditFraudBench (June 2026, work in progress) tests whether models can find where reported profit comes from, spot misleading narrative and map misconduct to known fraud patterns. GPT, DeepSeek and Qwen models, proprietary and open, struggled to reason jointly over figures, narrative and restatement evidence.
What it means: the hard parts of an audit, cross-document judgement and fraud reasoning, are not solved. Keep a person in charge of them, and test any tool on your own files. See AI hallucinations in audit for why errors matter.
Signal 4: In India, AI is now a professional-body priority
At its AI Innovation Summit in New Delhi on 26 June 2026, ICAI launched AICA Level 3, an advanced AI certification for members who have completed Level 2. It also signed an MoU with Sarvam AI, covering training, research and learning resources, and released a practice guide, AI Essentials for Chartered Accountants. ICAI said it had trained over 50,000 members and developed more than 150 GPT-based tools.
What it means: AI skills are becoming an expected part of a CA's toolkit. For how to read the certification path, see our review of AICA Levels 1 to 3 and the ICAI and Sarvam partnership explained.
Signal 5: What people are actually searching for
We pulled Google search volumes for India (OpenSEO, 16 October 2026) and compared the latest three months with the same months a year earlier. The numbers are small, so read the direction, not the decimals.
| Search | Monthly searches (recent) | Change vs a year earlier |
|---|---|---|
| ca gpt | about 22,000 to 33,000 | Peaked in July 2026 |
| ai for chartered accountants | about 450 | up about 53% |
| iso 42001 | about 4,400 | up about 61% |
| ai audit tools | about 100 | up about 68% |
| ai in audit | about 800 | flat |
What it means: curiosity about "AI in audit" as a general idea has levelled off. Interest is moving toward specific tools, professional upskilling and the governance standard for AI management systems.
Signal 6: Governance is catching up
Two things pull AI use toward governance. First, ISO/IEC 42001, the standard for AI management systems, is drawing growing search interest (see above), and we have a checklist for auditing against it. Second, India's DPDP Rules, notified on 14 November 2025, phase in over eighteen months, with most duties arriving around May 2027, which affects what client data can go into AI tools. See our DPDP audit readiness checklist.
What it means: "Which AI tool?" is quickly becoming "Where does the data go, and can we show our controls?"
Signal 7: Evidence and audit trail are becoming the buying criteria
Commentary on agentic AI points to the same blockers again and again: no audit trail, unclear decision lineage and unclear human oversight. For an auditor, the equivalent test is simple: can another reviewer repeat the work and see the source of every figure? We have a view here, since CORAA is built around a deterministic layer that gives the same output for the same input; see why we use deterministic AI and deterministic vs probabilistic AI. Judge any vendor, including us, on whether it can show its evidence.
What to do this quarter
- Write down your AI use policy: which tools, which data, who reviews.
- Run a small bake-off of two or three tools on your own anonymised files.
- Decide where client data may go, and check it against DPDP duties.
- Train the team. ICAI's AICA path is one route.
- Keep a log of AI-assisted work so a reviewer can reproduce it.
Frequently Asked Questions
What is trending in AI for audit in 2026?
Funding for audit-specific platforms, agent-style tools, audit-focused benchmarks that show model limits, professional-body programmes such as ICAI's AICA, and growing attention to AI governance standards like ISO 42001.
Is AI replacing auditors?
Nothing we read supports that. The benchmarks show models still struggle with cross-document reasoning and fraud judgement, so people stay responsible for conclusions.
What did ICAI announce on AI in 2026?
At its AI Innovation Summit on 26 June 2026, ICAI launched AICA Level 3, signed an MoU with Sarvam AI and released AI Essentials for Chartered Accountants.
Is there a binding standard on AI use in audits?
We did not find a binding AI-specific audit standard, but we could not open the PCAOB's pages and did not read IAASB material in full. Check the current position with the relevant standard-setter.
How should a CA firm start?
Set a use policy, test a few tools on anonymised files, decide where client data may go, and log AI-assisted work.
Related Articles
Statutory facts on this page are checked against their sources, and the page says where it relied on secondary reporting. How we verify · Report an error