What counts as a productive day

Productivity has been part of the Australian conversation for years. Economists worry about our lack of it, business owners ask how their teams can be more productive, and plenty of us finish a long day wondering whether it counted as a productive one. The word turns up in talk of an optimised life, in parenting, and around the table with family and friends.

The national numbers explain some of the worry. The Productivity Commission reported in September 2026 that Australia's labour productivity was flat in the June quarter and 0.2% lower over the year. Output grew 2.1%, and hours worked grew 2.3%.

AI has arrived in the middle of that conversation, along with the hope that it will make all of us more productive. In its February 2026 submission on productivity, the Reserve Bank wrote that investment in AI "may lift productivity over time, but the measured benefit of this is yet to be observed in Australia."

Before we can say whether AI is making us more productive, we need to know what we are measuring. What is productivity, and what counts as output in a law or accounting firm?

1Two kinds of productivity

Technically, productivity is output measured against input. The Australian Bureau of Statistics measures labour productivity as the value of what is produced for every hour worked. Economists call that value gross value added: the value of the goods and services produced, less the materials and services used up in making them, adjusted for changes in price. Add it up across every industry and you get close to GDP.

Two things follow from that definition. It counts hours, so if hours grow faster than output, as they did in the year to June, productivity falls. And it counts value, which is simple enough for a tonne of wheat and much harder for a piece of advice. The same letter of advice can carry years of experience or very little, and that difference is hard for any statistic to capture.

Law and accounting firms tend to measure productivity another way. In the Thomson Reuters Institute's benchmark of Australian law firms, productivity is measured in hours worked per fee earner, and it is reported alongside demand for legal work, the rates firms negotiate with clients, fees worked, and profit per equity partner.

Accounting firms have a similar habit. The Crunch, the annual benchmark report on Australian accounting firms, measures productivity as the share of a person's time that is chargeable. In its 2024 edition, partners in high-performing firms reported productivity of 50%, against 40% in lower-performing firms. Chargeable hours, recovery on those hours and fees per team member usually sit alongside it on a practice's dashboard.

All of these are measures of hours and dollars per person, and that creates an odd situation once AI starts saving time. The national measure improves when the same work takes fewer hours. By the firm's measure, a person who finishes the same work in fewer hours can look less productive, unless the time saved is filled with other chargeable work.

2Is AI making us more productive?

AI users say it saves them time, and how much depends on who is being counted.

• Across the American workforce, the time people report AI saving them came to 2.2% of all hours worked in May 2026. That figure includes the people who do not use AI, counted as saving nothing.

• In professional, scientific and technical services, the sector that takes in law and accounting, the same measure reached 3.94% of all hours in the June quarter of 2026.

• In a randomised trial across 66 firms, knowledge workers who used an AI tool built into their email and documents spent about two fewer hours a week on email.

Each of these counts something different: every worker, one sector, and the active users of one tool. A figure quoted without saying who it counts can make AI look far more useful, or far less, than the research shows.

What the saved time turns into is harder to see. Anders Humlum and Emilie Vestergaard linked a survey of around 25,000 Danish workers, in 11 occupations exposed to AI, to Statistics Denmark's records of what each of them earned and how many hours they worked. Their research found no measurable effect on either, and could rule out average effects on earnings larger than 2%. Most users, 85%, said they had moved the time into other job tasks. Of the new tasks AI had created, about 35% involved reviewing what the AI produced and making sure it complied.

In Australian law firms, the hours are already shifting. The Thomson Reuters Institute's February 2026 update reported "their third consecutive productivity decline" for non-equity partners and small falls for junior and mid-level associates, while the hours of senior associates and equity partners went up. The report called the evidence early and suggested AI as one possible reason. If AI is part of the reason, the lawyers whose research, drafting and review it takes on may be the ones who look less productive by the firm's own measure.

Accounting research shows the other side of an hours measure. Jung Ho Choi and Chloe Xie, writing in the Journal of Accounting Research in 2026, found accountants who leaned on AI more took on more clients and billed more hours, with less of their week going to data entry. Where saved time turns into chargeable time, an hours measure picks it up.

A survey of nearly 750 US executives by the Federal Reserve Banks of Atlanta and Richmond found companies believed AI had lifted output per worker by 1.8% in 2025. When the researchers worked the gain out from actual changes in revenue, it was much smaller across all major industries.

Some of what AI can add is quality, which sits outside an hours measure altogether. In one three-month trial with 133 practising patent lawyers at 11 US firms, work drafted with an AI assistant was rated higher by expert reviewers who did not know which drafts had used it, after ten days and again after three months.

If a firm's measure counts only hours per person, is it the right measure for the work firms do now?

3Output per client

Hours have measured firms well for decades, because for most of that time an hour of work produced a fairly predictable amount of output. For some tasks, AI is starting to change how much an hour can produce.

So what would it look like if a firm measured what its clients received, as well as the hours its people put in?

Many firms hold some of this already. Turnaround times can often be drawn from the practice management system. Write-offs show where recorded time and client value have come apart. Repeat work and referrals say something about whether clients valued what they got. And the team's hours, particularly outside normal working time, say something about what the output is costing the people producing it.

More output per person can come at a cost. A team can produce more files each and finish the year more tired for it.

4Deciding what counts

Picking from that table comes after a bigger question: what productivity should mean in the firm. A firm can work through it in five decisions.

And if a firm finds it is becoming more productive, a second decision is waiting: what the gain is for. It could go to taking on more clients, to more time with the clients the firm already has, or back to the team. We look at that choice in Working smarter with AI, not just faster and What to do with the time AI gives back. How the change shows up on an invoice is its own question, and Knowing where to tap takes it on.

5Where we come in

Our investigation measures how long the work your firm does most often takes today, so you start from a baseline of your own. The findings belong to your firm whether or not we go any further together.

If you would like to know what AI is changing in your firm, and how to measure it, reach out to our team. The first conversation works out whether your firm is the right fit for the AI Uplift Program, and whether we are the right people to help.

Deciding what counts