As EY rolls out an enterprise-scale multi-agent AI framework across 160,000 global engagements, the era of manual audit sampling is under threat. We analyze what a multibillion-dollar shift toward "agentic" AI means for audit quality, the rise of AI assurance, and the evolving role of the UK practitioner.
For the average UK practitioner, billion-dollar headlines about “agentic AI” often feel a world away from the reality of a Tuesday afternoon VAT return. However, EY’s global roll-out of enterprise-scale agentic AI is more than just a corporate flex. It is a signal that the fundamental unit of accounting, the “task” is being redefined.
The firm is embedding a new multi-agent framework, integrated with Microsoft Azure, Microsoft Foundry, and Microsoft Fabric, directly into its EY Canvas platform. This isn’t a pilot; it’s a deployment reaching 130,000 professionals across 160,000 audit engagements in over 150 countries.
If you aren’t at a Big Four firm, you might think this doesn’t affect you. In reality, it changes the benchmark for the entire profession. Here is a look at what this actually means for accountants on the ground.
For decades, the audit has been a game of statistical probability. We test a sample, find no errors, and gain “reasonable assurance.” This technology effectively seeks to end that era.
By processing 1.4 trillion lines of journal entry data per year through EY Canvas, the industry is moving toward 100% population testing. For the wider profession, this creates a “quality gap.” When mid-tier or small firms continue to use traditional sampling, they may soon face tougher questions from the FRC or professional indemnity insurers about why they didn’t catch an anomaly that a basic AI agent could have flagged in seconds.
EY’s commitment includes a global training program to upskill all audit and technology risk professionals this year. This isn’t just about learning new software; it’s a shift in the job description.
The Old Role: Data gatherer and manual “ticker.”
The New Role: AI Orchestrator. You aren’t doing the testing; you are reviewing the “logic” of the AI agent that did the testing.
The FRC is clear: human judgment remains the final word. If an AI agent misses a material misstatement because its “reasoning” was flawed, the human auditor remains legally accountable.
EY is positioning itself as “client zero,” testing these tools on its own business before rolling them out. Crucially, they are launching a new suite of AI assurance services spanning diagnostics, governance, and risk management.
As 97% of companies embark on their own AI transformations, they will need accountants to “audit the algorithm.” EY is building the infrastructure to do this at scale right now. If your firm isn’t at least exploring how to manage AI-related risks, you risk losing the high-value advisory ground to those who can prove they know how to govern a “black box” system.
One major risk that corporate announcements often gloss over is transparency. Agentic AI can be opaque; it doesn’t always show its “workings” in a way that a human can easily follow.
The Documentation Challenge: Auditors are trained to follow a trail. If an agent autonomously flags a risk without a clear “reasoning log,” documenting that for a file review becomes a nightmare.
The Hallucination Factor: Agents can occasionally “hallucinate” logic for instance, misinterpreting a complex lease agreement clause. Without “human-led, AI-powered” oversight, the risk of systemic error actually increases.
For the junior accountant who spent their first three years “pouching” invoices, this could be a massive win for mental health. It removes the “administrative burden” that leads to burnout.
However, for firm partners, the challenge is economic. Smaller firms must now decide whether to:
Partner: Lean on the AI tools being integrated into platforms like Xero, Iris, or Wolters Kluwer.
Specialize: Focus on high-touch, human-centric advisory where an AI agent lacks the emotional intelligence to navigate a family business dispute or a complex M&A negotiation.
The “experimental” phase of AI in accounting is over; we have entered the implementation phase. With EY aiming to support all end-to-end audit activities via AI by 2028, the “accountant of the future” must be a “power user” who understands the limitations, biases, and risks of the agents they oversee.
