Australia’s Treasury has published its 2026 Intergenerational Report, a 40-year projection that places the artificial-intelligence revolution alongside ageing, the care economy, geopolitical fragmentation, energy transition, industrial transformation and intergenerational equity as forces shaping the country through 2065-66. For business leaders, the significance is immediate: the report turns AI from a technology story into a productivity deadline.
The question readers actually need answered is not whether AI is impressive in demos. It is whether Australian companies can convert AI spending into measurable gains in output, quality and job design quickly enough to matter before demographic and budget pressures intensify. Treasury’s report makes clear why that matters. It does not settle whether companies will succeed.
A small productivity gap becomes a big fiscal story
Treasury’s central picture is still one of growth. It projects an economy more than twice its current size by 2065-66, with income per person 55 percent higher. But those gains sit on assumptions that are easy to state and hard to deliver inside real organisations.
One of the most important is long-run productivity growth of 1.2 percent a year, highlighted in Treasury’s accompanying remarks. Around that number, the budget arithmetic shifts fast. The underlying cash balance is projected to improve to a deficit of 0.3 percent of GDP in 2036-37, then widen to 1.8 percent by 2065-66. Gross debt is projected to fall to 22.2 percent of GDP in the mid-2050s before rising again to 27.4 percent by 2065-66.
That is why relatively small changes in productivity matter so much. Westpac said a long-run productivity rate 0.4 percentage points above Treasury’s assumption would cut gross debt by 25 percentage points of GDP in 2065-66, holding policy responses aside. The Business Council of Australia used the same report to press the opposite warning: recent labour productivity has averaged negative 0.1 percent a year over the past six years, and on its calculation a 0.8 percent long-run path would leave each Australian $20,700 worse off in 2065-66 than Treasury’s central case.
Those numbers are not a promise and not all from the same model. They do, however, sharpen the management challenge. If labour supply growth slows as the population ages, and if care demand rises, then output per worker matters more. Treasury expects participation to keep growing until 2040 and remain higher than previously thought, helped by women and older Australians. That buys time. It does not remove the need for firms to produce more value from the hours and capital they already use.
Why AI’s macro promise becomes a workplace problem
This is where the report’s AI section is most useful. Treasury describes AI as a defining influence on the next four decades because it can automate tasks, improve decision support, accelerate research and help create new products and services. All of that is plausible. None of it arrives automatically when a company buys licenses.
The gap between model capability and workplace performance is usually filled with unglamorous work: cleaning and governing data, redesigning processes, training staff, setting decision rights, checking cybersecurity, integrating systems and paying for the computing and energy needed to run them. A company can add an AI assistant to a workflow that still requires the same approvals, the same duplicate data entry and the same manual rework. In that case it has raised expense more quickly than output.
That is why the Intergenerational Report should be read less as a claim that AI will rescue the budget than as a stress test of corporate execution. Australia does not get the productivity upside from AI because the technology exists. It gets it if managers change how work is done, and if employees are equipped to use the systems well enough that quality holds up while cycle times fall.
Better jobs are part of that equation, not a side issue. If AI adoption simply shifts more checking, exception handling and uncertainty onto workers without giving them skills, authority or better tools, businesses may find that promised efficiencies are offset by higher error rates, employee resistance and churn. The report does not say AI will raise wages or wellbeing. Leaders need to make those outcomes part of the business case rather than assume they will appear.
What leaders should demand before scaling an AI program
The most useful response to the report is practical. Before scaling an AI program, executives should ask for evidence at the task and workflow level, not just vendor benchmarks or pilot anecdotes.
Start with the work itself. Which task is being automated or augmented? What is the baseline today for cycle time, throughput, error rates, rework and customer outcomes? What part of the process actually constrains output? If the bottleneck is approval, compliance review, missing data or a legacy system, an AI layer may do little beyond move the queue.
Then count the full cost of adoption. That includes integration work, model governance, security controls, staff training, supervision, change management and any added energy or computing demand. A program that saves minutes on one task but increases review requirements elsewhere can look efficient in a demonstration and uneconomic in production.
Finally, leaders should track what happens to people as closely as what happens to output. Are workers handling more complex cases because routine ones are automated? Are they spending less time on admin and more on revenue, care or problem-solving? Are new skills being built internally, or is the firm becoming dependent on imported models and outside contractors? The public material around the report does not answer those questions by occupation, sector or firm size. Boards and executives will have to.
What the report does not tell business yet
Treasury’s public summary elevates AI strategically, but it does not map which industries will capture the biggest gains, how quickly adoption will spread, which jobs will be augmented or displaced, or what implementation costs companies will bear. It also leaves open how much of the upside depends on domestic innovation versus imported models and infrastructure, and how added data-centre and energy demand will interact with the grid.
That uncertainty is not a flaw so much as a warning against easy extrapolation. The report sets the macro stakes. The next phase belongs to company operators: the retailers, banks, miners, manufacturers, health providers and professional-services firms deciding whether AI changes a process enough to lift real output. If they cannot show faster cycle times, fewer errors, higher throughput, lower rework and credible workforce transition plans, then AI remains an expense line attached to a national hope.
By
By
By
By
By
