AI's Double Edge: Canberra Businesses Weigh the Promise Against Real and Growing Risks

More than 60 percent of small and medium businesses in the ACT have now trialled at least one AI tool in their operations, according to a June 2026 survey by the Canberra Business Chamber — but fewer than a quarter say they have a formal policy governing how that technology is used. That gap is starting to matter.

The timing is pointed. Globally, regulators are tightening frameworks around AI deployment, and Australia's own AI Safety Standard — currently in its second public consultation phase, closing September 2026 — will require businesses above certain revenue thresholds to document risk assessments for high-stakes AI applications. For a city whose economy runs heavily on government contracts and professional services, the compliance question is not abstract. It lands on desks in Civic and Fyshwick alike.

The Local Picture: Enthusiasm and Exposure

At the Australian National University's 3A Institute in Acton, researchers have spent months documenting how organisations bolt AI onto existing workflows without interrogating the assumptions baked into the underlying models. The concern is not hypothetical. A recruitment firm operating out of Lonsdale Street, Braddon, confirmed to The Daily Canberra this week that it paused its AI-assisted resume-screening pilot after staff flagged that shortlists were systematically underrepresenting candidates from non-English-speaking backgrounds — a bias the vendor's own documentation had flagged as a known limitation, buried in section nine of a 40-page technical brief.

The ACT Government's Digital Ready for Business program, which has disbursed more than $2.8 million in grants since 2023, has begun requiring applicants to complete an AI ethics module before funding is approved for any AI-adjacent project. It is a modest guardrail, but coordinators at Access Canberra say uptake of the voluntary advanced module has doubled in the first half of 2026, suggesting the anxiety is real and operators want guidance rather than just permission.

Equally telling is what is happening at the newer co-working spaces. Entry Canberra, the startup hub on Mort Street, Braddon, runs a fortnightly AI clinic. Facilitators report that the questions have shifted. Twelve months ago, attendees mostly wanted to know which chatbot to use. Now the sessions are dominated by liability questions: who is responsible when an AI-generated client report contains a material error, and whether cyber insurance policies — many of which were written before large language models became mainstream — actually cover AI-induced data leakage.

The Ethical Weight Behind the Efficiency Pitch

The productivity case for AI remains compelling and should not be dismissed. Independent modelling by the Productivity Commission, published in March 2026, estimated that widespread AI adoption in professional services could lift sector output by 12 to 18 percent over five years. For a city where the professional services sector employs roughly 95,000 people, that is not a rounding error.

But the same modelling flagged concentration risk. Productivity gains tend to flow to firms that can afford the better tools — typically the larger consultancies anchored around the parliamentary triangle — while sole traders and micro-businesses in suburbs like Tuggeranong and Belconnen absorb the competitive pressure without the resources to match. Three local accountancy practices in Woden have already told clients they are moving to AI-assisted tax preparation for the 2026-27 financial year, reducing per-return costs by roughly $80. That is good news for the client, but it quietly removes junior work that has historically trained the next generation of accountants.

There is no clean resolution to any of this before the end of the financial quarter, or the financial year. Businesses that want to move carefully have practical steps available right now: check whether your AI vendor's model card or technical documentation discloses known bias benchmarks; review your cyber and professional indemnity policies with a broker who has read the new ISO 42001 AI management standard; and treat the ACT Government's Digital Ready ethics module as a floor, not a ceiling. The ANU's 3A Institute publishes open access guidance updated quarterly. The Canberra Innovation Network also runs an AI governance workshop series — the next session is scheduled for July 22 at Entrepreneur Square on Moore Street. Free, two hours, booking required. Start there.