Australia Just Published Its First Real AI Numbers

Qantas does not usually feature in AI commentary. This week it did, and for the most operator-shaped reason imaginable: the airline told the AFR that AI pattern-detection tools helped it reach an 86 per cent domestic on-time rate, its best in nearly a decade, and saved roughly A$30 million.

That story did not arrive alone. Over about ten days, a run of large Australian businesses put specific, quantified AI results on the public record. Taken together they form the first genuinely local evidence base for what this technology does inside a business, rather than what a vendor deck says it might do.

The case file

Qantas. A$30 million saved and the best domestic on-time performance in close to ten years, attributed to real-time AI pattern detection across operations (AFR, 8 May).

Herbert Smith Freehills. Contract delivery on a major matter cut from 28 days to six, with 200,000 documents culled to 18,500 and an 80 per cent time saving over manual review, using the Legora legal AI platform (AFR, 1 May).

WiseTech. Executive chairman Richard White told the Macquarie Australia Conference that a software upgrade which previously took a team of eight to ten people 244 hours was completed by one senior engineer in 15 hours using AI agents (AFR, 5 May).

Macquarie Bank. The bank reported 130,000 hours saved in seven months on Google's Gemini Enterprise, with about 80 per cent of its 5,000 staff using it daily (iTnews).

Deloitte. The firm is making a A$1 billion back-office bet on the expectation that AI will automate 30 per cent of consulting tasks within three years (AFR, 7 May).

And beneath the big names, NAB research reports 42 per cent of Australian small and mid-sized businesses already using AI, mostly in admin, marketing and decision support (Capital Brief).

What the numbers share

Look at where every one of those results came from. Not a moonshot, not a reinvented business model, not a chatbot on the website. Each one is a narrow, well-understood, high-volume internal process with a baseline the business was already measuring: on-time departures, document review days, engineering hours, staff hours on routine tasks.

That is the repeatable pattern. The wins are landing where three conditions hold. The process runs at volume, so small percentage gains compound into real dollars. The process was already measured, so improvement is provable rather than vibes. And the process sits inside the business, where you control the data and the workflow, rather than at the customer edge where the risk of a public mistake is highest.

The honest caveats

These are self-reported figures, and you should hold them the way you would hold any company talking about its own program. Macquarie's 130,000 hours comes with no published methodology. Qantas' A$30 million is the airline's own attribution. WiseTech's productivity number arrives alongside a plan to cut up to 2,000 roles, which is a reminder that these gains carry workforce consequences the headline numbers do not capture.

None of that makes the pattern wrong. It makes the pattern worth verifying in your own business rather than borrowing on faith.

Your version of this

A Qantas-sized program is not the point. The transferable lesson fits a 30-person business as well as a 30,000-person one.

Write down the three processes in your business that run at the highest volume. For each, ask whether you can state today's baseline in a number you would defend to your accountant. If yes, that process is a candidate. If no, the first project is measurement, which costs almost nothing.

Then run one candidate the way Freehills ran document review: a bounded piece of work, existing tooling, a before-and-after number, and a human checking the output. Six weeks is enough to know whether your version of the Qantas result exists.

The significance of this week is that "does AI work in an Australian business" now has published answers. The question left open is specific to you, and it is much cheaper to answer than it was a year ago.

Where does your business stand? The free AI Scorecard takes three minutes and shows you. If you want a straight steer from a person, book an AI Opportunity Call.