Australian government AI adoption is outpacing assurance, innovation consultancy warns

Australian government agencies are adopting artificial intelligence faster than they are building the capability to govern, operate and sustain it, according to innovation consultancy Disruptors Co.

The company today launched a dedicated public sector AI practice and an eight-stage AI Adoption Diagnostic designed to identify gaps between experimentation and operational readiness.

Australian National Audit Office data shows 56 public sector entities reported using AI during 2023-24. Of those, 36 — 64 per cent — had established internal policies governing its use. Only 15, or 27 per cent, had policies covering assurance over that use.

The gap is becoming more consequential as Commonwealth agencies implement the Australian Government's strengthened Policy for the responsible use of AI in government, including requirements for strategic direction, accountable officials, use-case owners, internal registers, staff training and impact assessments for higher-risk uses.

"The Australian public sector does not have an AI experimentation problem," said Gavin Heaton, Co-CEO of Disruptors Co.

"There are pilots, licences and use cases appearing across government. The harder question is whether agencies have the foundations to decide what should proceed, what should stop, and who remains accountable once a system is operating."

The diagnostic assesses organisational maturity across eight stages: Intent, Ground, Discovery, Proof, Capability, Production, Practice and Compound.

Each stage is scored across five maturity levels. Disruptors Co describes the transition into level four as "the cliff" — the point where agencies must turn promising proofs into governed, repeatable operating capability.

"A pilot can be delivered by a motivated executive, a skilled project team and a cooperative vendor," Heaton said.

"Production has to keep working after those people have moved on, when conditions change and when nobody is paying attention. That is where AI stops being a demonstration and becomes a public administration responsibility."

The eight scores also produce six organisational profiles designed to expose patterns that can be hidden by a single maturity score — such as advanced experimentation sitting on weak data foundations, or strong governance without workforce adoption.
Recent public audits illustrate the challenge.

A 2025 ANAO audit found the Australian Taxation Office had partly effective arrangements supporting AI adoption. At the time of the audit, 74 per cent of its production AI models did not have completed data ethics assessments and the agency lacked centralised visibility of all its AI uses. The ATO agreed to all seven ANAO recommendations.

The Australian Government's whole-of-government Microsoft 365 Copilot trial also found that while 77 per cent of surveyed participants were positive about the tool and wanted to continue using it, only one-third were using it daily. Training was most effective when tailored to employees' roles, agencies and operating contexts.

"Tool access can generate enthusiasm, but it does not automatically change practice," said Joanne Jacobs, Co-CEO of Disruptors Co.

"Capability is not knowing which button to press or how to write a prompt. It is knowing when an output can be trusted, what information can be used, when human review is required and when a use case should not proceed."

AI adoption in government requires a different standard

Disruptors Co says public sector AI requires particular attention because citizens often cannot choose an alternative provider when an automated system affects a government service or decision.

"A company that deploys a flawed AI tool may lose customers," Heaton said. "A government agency that deploys one can affect someone's payment, visa, licence, care or access to a service. That difference should shape the design from the beginning."
The new practice spans maturity diagnostics, use-case discovery and classification, governance frameworks, executive briefings, workforce capability, controlled proofs, production operating models and evaluation.

Its approach is based on capability transfer, with agency staff co-facilitating assessments and receiving frameworks and templates they can use independently.
"We measure success partly by how many decisions the agency can make without us," Jacobs said. "That includes projects it advances, projects it redesigns and projects it correctly decides not to build."

Most initial assessments require no access to agency data, and Disruptors Co does not use agency information to train or fine-tune AI models.

Further information: disruptorsco.com/ai-consulting/ai-consulting-services-for-government

ENDS
 

Notes to editors

Public data referenced in this release is drawn from:

Australian National Audit Office, Audits of the Financial Statements of Australian Government Entities for the Period Ended 30 June 2024

Australian National Audit Office, Governance of Artificial Intelligence at the Australian Taxation Office, February 2025

Digital Transformation Agency, Evaluation of the Whole-of-Government Trial of Microsoft 365 Copilot

Digital Transformation Agency, Policy for the responsible use of AI in government, Version 2.0

The AI Adoption Diagnostic and its six profile signatures were developed by Disruptors Co. See https://pathstate.disruptorsco.com/