Big firms push AI into audit

A Big Four accounting firm just put artificial intelligence agents into the middle of audit work that used to belong to junior staff and long checklists. Ernst & Young, known as EY, said on April 7, 2026 that it is rolling out enterprise-scale “agentic” artificial intelligence across its Assurance business and embedding it into EY Canvas, the firm’s global audit platform. (ey.com) EY says the system will support 160,000 audit engagements globally, and its Canvas platform already processes more than 1.4 trillion lines of journal-entry data each year. The new setup uses a multi-agent framework connected to Microsoft Azure, Microsoft Foundry, and Microsoft Fabric, which means the software is being built into the firm’s core workflow rather than used as a side tool. (ey.com) (cpapracticeadvisor.com) That changes what “doing an audit” looks like. Traditional audit work often involves selecting samples from huge piles of transactions, matching documents to those samples, and writing up evidence in standard formats; EY is now pitching artificial intelligence as a way to test much larger populations of data and move routine procedures faster. (accountancyage.com) (ey.com) The promise is simple enough to explain: instead of checking a few boxes from a warehouse, the software can help scan the whole warehouse. In audit terms, broader population testing can reduce the chance that unusual items stay hidden outside the sample, especially when companies generate millions of records across finance systems. (accountancyage.com) (ey.com) But EY is not selling this as a frictionless upgrade for its youngest employees. Business Insider reported on April 7, 2026 that Marc Jeschonneck, EY’s global assurance digital leader, said the tools will create a steep learning curve for junior audit staff even if they eventually make work easier. (businessinsider.com) (msn.com) That warning gets at the most important workforce shift in this story. For decades, entry-level auditors learned by doing repetitive tasks like tying numbers, tracing invoices, and documenting exceptions; if software takes over more of that groundwork, firms will need new ways to train judgment, skepticism, and review skills. (businessinsider.com) (icaew.com) EY is not moving in isolation. KPMG has already integrated generative artificial intelligence into KPMG Clara, its global smart-audit platform, and said in 2024 that the system would support about 90,000 auditors worldwide. (accountancyage.com) (microsoft.com) Deloitte has also been adding generative and agentic artificial intelligence features to Omnia, its audit platform. Deloitte said those upgrades include documentation review, drafting help, data extraction, and research support for its audit professionals, showing that the race is now about embedding artificial intelligence into the standard audit stack, not just experimenting in labs. (prnewswire.com) (internationalaccountingbulletin.com) PricewaterhouseCoopers, known as PwC, has been making the same bet from another angle. PwC has said it expects end-to-end artificial intelligence audit automation by 2026, and its own materials describe an “AI-native audit” in which artificial intelligence agents handle parts of risk assessment, testing, and compliance while human auditors keep responsibility for judgment. (scottishfinancialnews.com) (pwc.com.au) The other half of the story sits outside the audit firms themselves. Vertex, a compliance software company, said on April 7, 2026 that it is adding new artificial intelligence capabilities to Vertex Cloud so tax, finance, and information-technology teams can execute compliance work faster while keeping accuracy, accountability, and audit readiness. (vertexinc.com) (cpapracticeadvisor.com) Vertex’s language is revealing because it focuses on “explainable and reviewable outcomes” and on reducing dependence on specialized skills through plain-language task completion. That is exactly where audit and compliance are heading: if more work is produced by artificial intelligence, firms will need cleaner records showing what the system did, what data it used, who reviewed it, and why the final answer was accepted. (vertexinc.com) (cpapracticeadvisor.com) That is why the next fight is not really about whether artificial intelligence can draft a memo or flag an outlier. The harder problem is governance: traceability, human review, model controls, data lineage, and evidence that an artificial intelligence-assisted conclusion can survive regulator, inspector, and audit-committee scrutiny. (pwc.com) (pwc.co.uk) (deloitte.com) EY’s announcement matters because it turns that future into current operating reality at large scale. Once one of the biggest audit networks in the world says agentic artificial intelligence will sit inside daily assurance work, clients