FY26 accounting firm benchmarks, and how to read your own

Every owner wants to know the same thing at the end of a financial year: was that a good year, or did it just feel busy? FY26 is a harder year to answer that for than most. AI has started changing how many hours a job takes, which quietly breaks the measures most firms use to judge themselves.

This piece collects the first published 2026 benchmark numbers we could verify for small and mid-sized firms in the US, Australia and the UK, says what each one actually measures, and sets out how to read your own FY26 numbers against them. As at 9 November 2026, there is no single published FY26 benchmark for small firms across all three markets, so treat what follows as reference points, not a league table.

It depends where you are. For an Australian firm, FY26 ran from 1 July 2025 to 30 June 2026. Most US CPA firms run on the calendar year, so the closest US equivalent is calendar 2025, which is what the 2026 US surveys report on. UK practices use a range of year ends. When you compare yourself with a survey, check the period it covers before you compare anything else.

The most detailed 2026 numbers published so far come from the 2026 Rosenberg Survey, now in its 28th year and compiled by The Growth Partnership. It covers 260 participating CPA firms and reports on 2025 results. The figures below are from Rosenberg Associates’ summary of 23 September 2026; the full report is sold separately. The survey skews towards firms well above ten people, with many headline figures quoted for firms over two million dollars in revenue.

Show the numbers

Period	Value
2023	10.7%
2024	10.7%
2025	7.9%
2026	7.6%

Source: Rosenberg Associates, 2026 Rosenberg Survey: 7 Things You Need to Know, September 2026

Growth slowed from the 10.7% high-water marks of the 2023 and 2024 editions, and the summary says 7.0 points of the 7.6% was organic. Four other figures from the same summary are worth knowing:

• Revenue per FTE averaged $222,000 for firms over two million dollars in revenue, up from $210,000 in the 2025 survey.
• Income per partner averaged $640,000, up from $615,000. Firms under five million dollars went the other way, falling to $359k from $463k.
• Billable hours for firms under twenty million dollars averaged between 1,345 and 1,366 a year.
• Fewer staff are logging high hours. In 2021, 30% of staff logged fewer than 1,400 billable hours a year. In 2026 that figure is 62%.

That last line is the one to sit with. Billable hours per person are falling while revenue per person rises. Some of that will be policy (firms deliberately reducing hours to keep staff) and some will be technology. The survey summary does not separate the two, and neither should you assume it.

For small Australian firms, the dedicated benchmark is The Crunch from The Benchmarking Group, which we covered when 2026 participation opened. Its results go to participating firms, and we could not find published headline figures from the 2026 edition as at November 2026.

The largest published Australian practice survey of 2026 is Xero’s Modern Practice Playbook, released on 29 September 2026 from 524 senior accountants and bookkeepers. It was commissioned by Xero, with respondents drawn from third-party panels and Xero’s own customer base, so it describes Xero practices more than the whole profession. Its figures, exactly as published:

• 85% of practices have tried AI; 52% use it day to day, rising to 69% among top-performing firms.
• 35% said AI has let them maintain or grow output without expanding headcount.
• Firms bill an average of 69% of client engagements on non-hourly terms.
• 88% of practices do not expect AI to reduce team numbers over the next year.

We did not find an independent 2026 benchmark for UK practices of ten to thirty people. Xero’s UK edition of the same playbook, published 4 August 2026 from 520 UK practices surveyed between 7 and 26 May 2026, reported that AI saves practices 7.1 hours a week on average, that advisory carries the highest margin of any service at 51%, and that only 5% of UK practices expect AI to reduce headcount within the next year. The time savings are self-reported. TaxCalc’s annual Accountants’ Voice survey had not published 2026 results when we checked.

Outside surveys tell you what a group of other firms did. Your own numbers, calculated the same way every year, tell you what changed in yours. Our nine practice benchmarks explains how to calculate each one from the ledger and practice management system you already run. What follows is how AI changes the reading in FY26.

Revenue per FTE: the headline, if you calculate it honestly

This is the single best measure of whether AI is paying off, because it does not depend on recording time. Include yourself and any offshore or contract staff doing client work, or the number flatters you. Compare FY26 with FY25 and FY24, not with a survey average from a different-sized firm.

Utilisation: expect it to fall, and decide if that is good

If AI shortens jobs and you have not added clients, chargeable hours fall and utilisation drops. That is not failure. It is capacity. The question is whether you used it (new clients, new advisory work, fewer late nights) or whether it leaked into internal tinkering. Read it next to revenue per FTE: utilisation down with revenue per FTE up is the shape you want.

Realisation and write-offs: watch for the hourly leak

Firms still billing hourly will see realisation look healthy while revenue per job falls, because there are fewer hours to bill. Firms on fixed fees should see effective return per hour rise. If it did not, find out where the saved time went.

Wage-to-revenue ratio: the AI test

If AI is saving the time it promises, revenue should have grown faster than payroll in FY26. If the ratio rose, either the tools are not saving time on the work that matters, or the saved time is not reaching the invoice. Add your AI subscriptions to the cost side so the comparison is fair.

Lockup days: the number AI does not fix

Work done faster still has to be billed and collected. If WIP and debtor days did not fall, the gain is sitting in your balance sheet, not your bank.

“Inside the Numbers: What FY26 Really Looked Like for 250+ Firms” with Ryan Richardson of Dashboard Insights is on the agenda at The Firm’s AI in Practice Summit, 11 and 12 November 2026. We will add his figures to this piece after the session, with his permission; until then, we are not going to guess at them. Notes from the summit will be at thefirm.media/summit/november-2026/notes.

This is also the first step towards an annual State of the Firm report, which The Firm will launch in 2027, an annual look at how firms actually performed; no date is set yet. Until it exists, the most useful benchmark you have is your own FY24, FY25 and FY26, side by side. Calculate your own numbers before you compare them with anyone else’s.

The assumption most likely to mislead is that these surveys describe a firm like yours. The Rosenberg figures come from US CPA firms, mostly above two million dollars in revenue; the Xero figures come from a vendor-commissioned survey that partly drew on Xero’s own customers, and its time and money savings are self-reported estimates. None of them isolates what AI did from what pay rises, hiring shortages and price increases did. We read only the published summaries, not the full paid reports, and we have not been able to verify any 2026 results from The Crunch.

Is there an FY26 benchmark for small Australian firms?

The Crunch is the dedicated one, and its results go to participating firms. Published 2026 Australian figures so far come from Xero’s practice survey, which is broader and vendor-commissioned.

Why does the US survey call it 2026 when it covers 2025?

Surveys are named for the year they are published. The 2026 Rosenberg Survey reports on 2025 results, which for a calendar-year US firm is the year that overlaps the second half of an Australian FY26.

Should I compare my revenue per FTE with the survey average?

Only loosely. The US average is for firms over two million dollars, in a different market and currency. Your own trend over three years is more useful than any outside average.

My utilisation fell in FY26. Should I worry?

Not by itself. If revenue per FTE held or rose, falling utilisation means you freed capacity. Worry if both fell, because that means the saved time went somewhere that does not pay.

When will The Firm publish its own benchmarks?

The State of the Firm report launches in 2027; no date is set yet. In the meantime we will add the FY26 figures from the November summit to this piece once we have permission to publish them.