How the Big 4 Are Using AI and Automation to Reinvent Audit and Tax

Something fundamental has shifted inside the world's largest accounting firms, and it is not subtle.

Walk into a Deloitte audit team today and you might notice something strange: the junior associate is not knee-deep in a spreadsheet testing samples. She is reviewing exceptions that an AI model has already flagged, asking the question only a trained professional can answer. Why did this happen, and does it matter?

That shift is from doing the work to evaluating the work is the clearest signal of where the Big 4 are headed. Deloitte, PwC, EY, and KPMG are not just dabbling in artificial intelligence. They are rebuilding their operating models around it. And the implications for audit, tax, and every professional who works in these functions are more significant than most conversations acknowledge.

From Sampling to Full-Population Testing

For decades, one of the most fundamental constraints in auditing was sample size. Auditors could not realistically examine every single transaction a large company processed in a financial year. So they sampled, selected a representative subset, tested it rigorously, and drew conclusions about the broader population.

It was a reasonable approach. It was also an inherent limitation.

AI has changed that equation entirely. Today, firms like EY are analysing 100 percent of client journal entries through their audit analytics platform, EY Helix are not samples, but entire populations of data. KPMG's Ignite platform uses machine learning to scan millions of accounting entries, surfacing anomalies for human review. PwC developed GL.ai, a general ledger analysis tool built in collaboration with H2O.ai, designed to catch irregularities that would be invisible to even the most diligent human reviewer working through a traditional sample.

What this means in practice is significant. The audit is no longer a backward-looking exercise constrained by what a team of five could manually review in six weeks. It is a continuous, data-driven process capable of examining the full financial record of a business and identifying the specific transactions that warrant closer human scrutiny.

The auditor's role does not disappear. It evolves. Judgment, professional scepticism, and communication, the parts of auditing that require a trained human mind that become more prominent, not less, when the routine work is handled by an algorithm.

Evolution of Agentic AI, What It Actually Does?

In 2025, all four firms moved beyond traditional automation into what the industry calls agentic AI, systems that do not just assist with tasks but complete them, end to end, with minimal human intervention.

EY armed 80,000 of its tax professionals with access to 150 AI agents through its EY.ai Agentic Platform. These agents handle routine compliance tasks, process client data, and help professionals navigate the kind of regulatory complexity that once required hours of manual research. According to EY's own data, the platform processes millions of compliance cases annually. The pitch from EY's leadership is direct: professionals equipped with this technology can serve more clients, with greater depth, than they ever could working alone.

Deloitte took a different architectural approach with Zora AI, its agentic platform launched in early 2025 and built in partnership with Nvidia. Zora is designed to automate finance and procurement workflows such as invoice processing, trend analysis, variance reporting — tasks that previously consumed significant capacity across finance teams. Deloitte projects that its finance agents could free up thousands of working hours annually and reduce costs in these functions by as much as 25 percent.

KPMG's answer is Workbench, launched mid-2025, which takes a multi-agent approach. Rather than deploying a single AI system to complete a task, Workbench orchestrates multiple agents working together, mirroring how a human audit team operates, with different members responsible for different aspects of the engagement. The platform builds in documentation and accountability logs at every step, addressing one of the most persistent concerns about AI in regulated environments: explainability.

PwC developed Agent OS, its AI operating system, with governance and compliance at its core. The platform is modular so the firms and clients can configure it for their specific regulatory environments and is designed to appeal particularly to industries where errors carry serious consequences: banking, healthcare, infrastructure. The emphasis on governance is deliberate. In audit and tax, an AI tool that cannot explain its reasoning is not just unhelpful, it is a liability.

Tax: The Function AI Is Transforming Fastest

If audit is changing, tax is being restructured from the ground up.

Tax work has always been a peculiar combination of the highly mechanical and the intensely technical. On one end: processing returns, reconciling figures, checking calculations against filing requirements. On the other end: interpreting regulatory changes, advising on structuring decisions, defending client positions to revenue authorities. AI is extraordinarily capable at the first category and entirely incapable of the second.

The Big 4 recognised this split early and built around it.

Adoption of AI-assisted tax preparation tools surged dramatically between 2024 and 2025, with some firms reporting that over 80 percent of individual return preparation is now handled through automated workflows. What would once have occupied a junior associate for a full day — gathering documents, checking entries, generating the return, is now completed in a fraction of the time. Large language model-based research tools are reducing the time professionals spend on document analysis by more than half at several firms.

KPMG has invested $2 billion in cloud and AI capabilities, including a significant Microsoft partnership, with a declared goal of generating $12 billion in additional revenue from AI-enabled services. The strategy is not simply to do existing tax work faster so, it is to deliver categories of insight that were previously too expensive or too time-intensive to offer at scale.

EY's approach to tax AI is particularly transparent about its ambitions. The firm has described a "service-as-software" model, a future in which certain tax services are delivered through technology platforms, with human professionals engaged for the judgment calls that fall outside what any model can reliably handle. This is not a distant aspiration. It is the direction the current investments are pointed.

What AI Adoption Is Doing to the Profession

The impact on talent, hiring, and career paths inside the Big 4 is real and worth addressing directly.

Entry-level hiring patterns are already shifting. A 2025 Stanford study found that hiring for junior accounting roles fell by 16 percent over roughly two years, a direct consequence of automation absorbing the routine work those roles traditionally handled. Firms are not hiding this. They are describing it as a structural shift toward teams that are smaller, more senior, and more advisory in character.

The firms that are managing this transition well are doing two things simultaneously: deploying AI to raise the productivity ceiling, and investing heavily in reskilling the people who remain. EY's internal upskilling initiative has trained more than 55,000 employees in AI-related capabilities. KPMG reports that a continuous AI training programme led to measurable gains in staff confidence and adoption. The message to the profession is consistent across all four firms: the professionals who thrive are those who understand how to work alongside AI, what to delegate to it, when to trust it, and when to override it.

The broader adoption numbers tell a story of rapid movement. According to the 2025 Wolters Kluwer Future Ready Accountant report, AI adoption across accounting firms jumped from 9 percent in 2024 to 41 percent in 2025. That is not incremental change. That is an industry crossing a threshold.

The Risks Nobody Is Talking About Loudly Enough

The Big 4's investment in AI is real, the tools are functioning, and the efficiency gains are measurable. But there are genuine risks in this shift that deserve more prominent attention than they typically receive.

The first is the explainability problem. Audit opinions carry legal weight. When a machine flags an anomaly or clears a transaction, someone needs to be able to explain, to a regulator or a court, exactly why that conclusion was reached. Documentation and audit trails within AI systems are improving — KPMG's Workbench is specifically designed around this — but the question of ultimate accountability has not been fully resolved by any firm.

The second is the skills pipeline. If entry-level roles are declining and junior professionals no longer spend years processing transactions, learning the mechanics of financial data from the ground up, what happens to the quality of the senior judgment that depends on that experiential foundation? It is a question the profession is only beginning to confront.

The third is the commoditisation risk. KPMG itself demonstrated this when it used its own AI adoption as a justification for negotiating a lower audit fee with its external auditor on successfully reducing costs by 14 percent. If AI makes audit and tax services faster and cheaper to deliver, clients will eventually expect those savings to be passed on. The firms that built their economics on billable hours are navigating toward a pricing model they have not fully defined yet.

Where This Is Heading?

The trajectory is clear, even if the destination is not perfectly mapped.

The Big 4 are not becoming technology companies. They are becoming professional services firms that use technology as infrastructure — the way they once used analytical spreadsheets or statistical sampling software, but at a scale and a speed that changes what is possible in fundamental ways.

The audit of the future tests everything, continuously. The tax function of the future processes compliance automatically and focuses human expertise on interpretation and advisory. The professional of the future is fluent in both finance and AI, not as a programmer, but as someone who understands what the tools can do, what they cannot, and how to take responsibility for the judgment calls that sit beyond the reach of any algorithm.

The firms that figure out how to combine the credibility of deep human expertise with the scale and consistency of AI will be formidable. The ones that over-rotate toward automation and hollow out the judgment underneath it will find, eventually, that what they are selling is a great deal less valuable than they think.

The technology is not the story. How these firms use it, and what kind of professionals they build around it is.

The Big 4's AI transformation is moving faster than most professionals in the field realise. Staying current on how these tools are reshaping audit and tax is no longer optional, it is a career imperative.