Zato hails AI agents for accountants amid staff squeeze

Zato has published two whitepapers on multi-agent artificial intelligence for accounting, tax and audit work in Australia and New Zealand, addressing a regional shortage of qualified accountants.

The New Zealand company argues that teams of specialised AI agents, operating under human oversight, are better suited to compliance and professional services work than single general-purpose models. One paper cites research showing collaborative multi-agent systems can improve accuracy by up to 40 per cent in complex analytical tasks.

The publication comes as accounting firms across Australia and New Zealand face mounting hiring pressure and heavier regulatory workloads. Drawing on data from Chartered Accountants Australia and New Zealand, the papers say more than 90 per cent of firms are struggling to recruit skilled accountants.

In Australia, enrolments in the Accounting Professional Year programme fell from 7,122 in 2018 to 340 in 2024, a drop of more than 95 per cent. The papers also cite Australian Bureau of Statistics projections showing demand for 338,362 accountants by 2026, requiring nearly 10,000 new professionals each year, while about 22,000 experienced practitioners are expected to retire over the same period.

That gap is colliding with an expanding compliance burden. In New Zealand, Inland Revenue received an additional NZ$116 million in enforcement funding in 2024 and aims to return NZ$4 for every extra dollar spent. Its interpretation and consultation notes rose from five across the previous four years to 22 in 2024 alone.

Indirect tax and reporting requirements are also becoming harder for firms to manage. Inland Revenue has identified 227,000 crypto-asset users conducting NZ$7.8 billion in transactions, while climate reporting and assurance requirements are adding work for Australian auditors.

Dual-zone model

Zato's research outlines what it calls a dual-zone architecture that separates routine preparation from work requiring professional judgment. Under the model, an Autonomous Execution Zone handles 11 categories of preparatory tasks, including data gathering, document normalisation, transaction classification, reconciliation, workpaper drafting and anomaly detection.

A separate Governed Decision Zone is reserved for judgment-based and legally consequential work, such as forming opinions, regulatory submissions and client advice. The papers describe a four-stage framework in which autonomy progresses through Shadow, Assisted, Supervised and Constrained modes.

The second paper focuses on Zato's Ziffy engine, which coordinates seven specialist agents covering tax, audit, compliance, financial analysis, GST and BAS, payroll, and strategic advisory. Zato said the structure is designed to keep outputs as analytical support rather than formal advice.

According to the research, large firms have also been moving towards multi-agent systems. The papers cite EY's use of 150 AI agents to support 80,000 tax professionals and refer to similar programmes at KPMG and Deloitte, while Gartner reported a 1,445 per cent rise in inquiries about multi-agent systems between the first quarter of 2024 and the second quarter of 2025.

Client demand for firms using more advanced tools may also be strengthening. Zato cited a BDO Audit Innovation Survey that found 97 per cent of finance leaders were willing to pay more for audit firms using advanced technologies, while 81 per cent said they had greater trust in technology-forward firms, up 18 percentage points year on year.

For Zato, the issue is that staffing shortages have become too severe for firms to solve through recruitment alone.

"The profession does not have a recruitment problem it can hire its way out of. The pipeline has collapsed faster than firms can replace it," said Saurav J Bansal, Chief Executive of Zato. "Our research is not about replacing accountants. It is about giving the ones who remain a way to absorb rising regulatory volume without compromising the standards their clients and regulators depend on."

Zato argues that the distinction between routine processing and professional judgment is central to how such systems should be used in regulated work.

"A reconciliation is not an audit opinion, and the system should never treat them the same way," said Dr Srinivas Kishan Anapu, Executive Director of Zato. "The dual-zone model makes that distinction explicit and enforceable, so routine preparation can move at machine speed while every judgment that carries professional or regulatory weight stays under human control, with a full reasoning trace behind it."