Australia’s business use of artificial intelligence has accelerated quickly. The Australian Bureau of Statistics reports that 12 per cent of businesses used AI in 2024–25, up from 1 per cent in 2022–23. That jump gives policymakers and executives a reason for optimism. It also creates a practical question that rarely appears in adoption statistics: who checks whether AI improves the work after employees start using it?

In most organisations, that responsibility lands with managers. They decide which tasks can use AI, review unusual outputs, settle disputes about accuracy, and explain new expectations to employees. Yet many AI programmes measure licences, training completion, or tool usage while overlooking the manager’s growing burden of verification, correction, escalation, and reassurance.

Australia can get more dependable productivity from AI by making a manager audit part of every rollout.

The audit should start with one real work process rather than a broad technology survey. A manager can select a recurring task such as preparing a customer response, summarising a case file, drafting a quotation, or forecasting inventory. For two weeks, the team records six facts: which tool produced the output, where a person checked it, how much time the tool saved, how much time people spent correcting it, what exceptions required escalation, and whether the result created extra work for someone downstream.

This record changes the conversation. A tool that saves forty minutes in drafting but creates thirty minutes of checking and ten minutes of customer recovery has delivered little dependable value. Another tool may save only fifteen minutes but produce consistent work with clear human review. The second result can support safer scaling even though the headline time saving looks smaller.

The Australian Government’s Guidance for AI Adoption already emphasises accountability, risk management, transparency, and human oversight. A manager audit turns those principles into daily operating evidence. It identifies who owns a decision, when a person must intervene, and which effects deserve attention before an experiment becomes standard practice.

Managers can use a simple traffic-light decision after the audit. Green uses have low correction rates, clear ownership, and no unresolved customer or worker harm. Amber uses show value but need tighter instructions, better data, or more training. Red uses create repeated errors, unclear accountability, or consequences that the team cannot reliably reverse. The organisation can scale green uses, repair amber uses, and pause red uses.

Frontline participation matters. Employees often notice awkward workarounds before senior leaders see them. They know when a summary leaves out a critical detail, when a recommendation does not fit the customer, or when a supposedly faster process simply shifts work to another team. A short weekly review gives those observations a formal route into deployment decisions.

Jobs and Skills Australia’s transition study describes AI adoption as a multi-speed process across industries and occupations. That unevenness makes local managerial judgment especially valuable. A national strategy can set direction, but each workplace still needs evidence about its own tasks, risks, skills, and customers.

The goal should be net dependable work: useful output that remains after checking, correction, rework, and recovery. Measuring that result protects productivity from inflated claims and protects employees from being blamed for problems created by weak implementation.

Australia has moved beyond asking whether businesses will use AI. The next question is whether organisations can learn from its use quickly enough to improve it. A manager audit offers a modest, repeatable answer. It gives leaders a clearer productivity measure, gives workers a voice in correcting the system, and gives customers a better chance of receiving work that someone can stand behind.

---

Bio:
Gleb Tsipursky, PhD, a behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026). https://disasteravoidanceexperts.com/aibook
Author contact: gleb@disasteravoidanceexperts.com