Because saved hours only become profit when a leader decides where they go. McKinsey's State of AI 2026 survey found 80 per cent of respondents say AI has improved their productivity, whilst only 6 per cent of organisations attribute an EBIT improvement of 5 per cent or more to AI (McKinsey, 2026). The gap is a management decision, not a technology problem.
The idea
AI is producing genuine individual time savings and almost no measurable earnings impact, and the variable that separates the two outcomes is whether somebody redesigned the work. McKinsey's State of AI 2026 survey, run between 4 May and 8 June 2026 across 1,719 respondents in 97 countries, found 37 per cent of organisations attribute at least some EBIT impact to AI and only 6 per cent attribute 5 per cent or more (McKinsey, 2026). Roughly three quarters of that 6 per cent had fundamentally redesigned their workflows, against about a quarter of everyone else. Redesigning the work also changes who does the foundational tasks, which is the same pipeline question as where your next senior hire comes from.
Both figures are global rather than Australian, and the Australian read-across is that the time savings are real here too. Research by Agile Market Intelligence for The Access Group, covering 434 Australian accountants and bookkeepers, found 37 per cent saving at least 30 minutes a day and 56 per cent reporting productivity gains since adopting AI (The Access Group, 2025). Australian adoption is climbing at the same time, with the Australian Bureau of Statistics reporting that 12 per cent of all Australian businesses used AI in 2024-25, rising to 22 per cent of medium businesses and 35 per cent of large ones (ABS, released 25 June 2026), and the National AI Centre reporting 43 per cent of Australian SMEs with some level of AI adoption over the December 2025 to February 2026 quarter (National AI Centre, 2026).
Why it matters
For an Australian owner, the commercial consequence is that AI licence spend is already sitting on the profit and loss whilst the return sits in hours nobody has claimed. Boston Consulting Group's 2026 AI at Work survey of nearly 12,000 frontline employees, managers and leaders across more than a dozen markets found that 42 per cent of regular AI users save at least eight hours a week, and that 66 per cent of employees receive limited or no guidance on what to do with the time AI frees up (BCG, 2026). More than half do not reinvest that time in higher-value work. The unallocated hours problem is the same one that sits behind how much of an Australian broker's day actually goes on advice, and it is the reverse of where AI is moving construction margin, where the recovered hours were assigned to a named task from the start.
Worth flagging that the same BCG research found only 36 per cent of employees felt they had been adequately upskilled whilst 72 per cent said the skills expected of them had shifted, and only 33 per cent said leadership communications about AI were clear (BCG, 2026). The instruction layer is missing, not the tooling layer. The scale of what is being left on the table is not trivial either, with EY-Parthenon modelling AI as worth between $95 billion and $116 billion to the Australian economy by 2036 on a 2.0 to 2.4 per cent lift in multifactor productivity (EY, 2026).
Here is the difference in practice between a business getting nothing and a business in the 6 per cent:
| The step | What most businesses do now | What the 6 per cent do |
|---|---|---|
| Rollout | Buy licences for the whole team, run one introductory session | Pick one workflow, name its owner, record a before figure |
| Saved hours | Absorbed back into the working day | Assigned in writing to a named activity |
| Workflow design | Steps stay the same, with AI dropped inside them | The manager who owns the workflow rewrites the sequence |
| Measurement | Licence uptake and staff sentiment | Cost per file, hours per job, or conversion rate |
| Result | Individual productivity, no movement in EBIT | 5 per cent or more of EBIT attributed to AI |
The Alvo take
The hours are real, they are just unallocated, and an owner can size the problem in about ten minutes with numbers already on hand. Take a 20-person Australian business where 12 people use AI regularly. Applying the BCG finding that 42 per cent of regular users save at least eight hours a week gives you five people saving eight hours each, which is 40 hours a week, or one full-time equivalent that has quietly appeared inside the business.
Run that out across a working year. Forty hours a week over 46 working weeks is 1,840 hours, and at a fully loaded cost of $75 an hour (swap in your own figure) that is $138,000 of recovered capacity. BCG found more than half of employees do not reinvest saved time in higher-value work, so on those inputs roughly $69,000 of it has no destination this year. Nobody stole it, nobody was asked what to do with it.
The pathway out is narrow and unglamorous. Pick one workflow, quoting, file preparation, candidate screening, whichever one your team complains about most, and measure what it costs today in hours and cost per file. That measurement is a structured assessment of your workflows rather than a software decision. Redesign the sequence rather than inserting AI into the existing steps, because a quoting process that still routes every draft through the same two people has not been redesigned regardless of what software sits inside it. Then write down where the recovered hours go, whether that is more client contact, a role you no longer need to backfill, or a contractor you stop renewing. That last step is the one almost everybody skips, and it is the only one that reaches the P&L.
AI is producing real individual time savings and almost no measurable earnings impact, and the variable that separates the two is whether a leader redesigned the work and named where the hours go. On BCG's numbers a 20-person business with 12 regular AI users has one full-time equivalent of recovered capacity, worth around $138,000 a year at $75 an hour, and roughly half of it has no destination. The fix is one workflow, one before figure, and a written decision about what the recovered hours are for.
Sources: McKinsey & Company, The State of AI: Global Survey 2026, published August 2026. Boston Consulting Group, AI at Work: Why Strategy Matters More Than Tools, June 2026. Australian Bureau of Statistics, Business Adoption of Artificial Intelligence 2024-25, released June 2026. National AI Centre, AI adoption insights December 2025 to February 2026, May 2026. EY-Parthenon, AI productivity gains could deliver up to $116bn boost to Australia's economy, August 2026. The Access Group, research by Agile Market Intelligence, 434 Australian accounting professionals, fieldwork September to October 2025.
Common questions
Why is AI not showing up in our profit even though our staff say it saves them time?
Saved hours only reach the profit line when a leader reassigns them. BCG's 2026 AI at Work survey found 66 per cent of employees receive limited or no guidance on what to do with the time AI frees up, and more than half do not reinvest it in higher-value work. Without that instruction, the hours are absorbed back into the working day and nothing changes in the numbers.
How do I measure the return on AI in an Australian SME?
Measure the destination of the hours, not the hours themselves. Pick one workflow, record the time it took before AI and after, then name in writing what the recovered hours will be used for, whether that is more client contact, faster quoting or one role you no longer need to backfill. The return shows up in that second number, not the first.
What does redesigning a workflow actually mean in a small business?
Redesigning a workflow means changing who does which step and in what order, not adding AI to the existing steps. A quoting process that still routes every draft through the same two people has not been redesigned, whatever software sits inside it. McKinsey's 2026 survey found roughly three quarters of the organisations getting real earnings impact from AI had rebuilt their workflows this way.
Should we roll AI out to the whole team or start with one department?
Start with one workflow in one team, because a whole-of-business rollout multiplies licence cost before anyone has proven where the hours go. Australian Bureau of Statistics data shows 22 per cent of medium-sized Australian businesses used AI in 2024-25, so most owners are still early enough to sequence this deliberately. One workflow gives you a measurable before and after you can use to justify the next one.
How long does it take before AI shows up in the P&L?
Expect two to three months for a single redesigned workflow, and longer if the saved hours have to convert into revenue rather than cost. Cost effects land faster because a role left unfilled or a contractor not renewed shows immediately. Revenue effects lag because more client contact takes a sales cycle to appear as settled or invoiced work.
Is AI training for staff worth paying for?
Training is worth paying for when it is tied to a specific workflow rather than delivered as general AI literacy. BCG's 2026 survey found only 36 per cent of employees felt they had received adequate upskilling, whilst 72 per cent said the skills expected of them had shifted. A half-day session on the quoting process your team runs every week will outperform a generic AI course.