Thursday, 6 August 2026 • By Timothy, CPA — Managing Director, Professional Financelink (PFL)

Stylised illustration of two diverging growth lines emerging from a single point, with a magnifying glass hovering over the assumption at their origin

AI Could Add $4.8 Billion a Year to Australian Accounting — Read the Assumption Before You Read the Number

New MYOB modelling puts a big figure on AI adoption. The headline is a forecast built on one assumption, and that assumption is the most useful thing in the report.

On 30 July, MYOB published modelling estimating that wider use of AI and automation could add AUD $4.8 billion a year to Australia's small and local accounting sector within five years, starting at $835 million in additional annual revenue in year one. The forecast covers established small, local and solo practices — excluding the Big Four and national mid-tier firms — which MYOB says represent the majority of Australia's roughly 37,000 accounting practices.

It's a good number for a headline. It's a less good number for a decision. And the reason is sitting in plain sight in the methodology, which is genuinely worth more attention than the total.

What the Modelling Actually Does

The model compares two groups. Practices that don't adopt AI are assumed to grow revenue by 1.7 per cent a year. Practices using paid AI subscriptions are assumed to grow at 5.3 per cent. The gap compounds across five years, applied across the segment at full adoption, holding the number and mix of practices constant.

Note the verb. Assumed. The 1.7 and the 5.3 are inputs to the model, not outputs of it. Everything downstream — the $835 million, the $4.8 billion — is arithmetic performed on those two numbers. If the growth differential is 3.6 percentage points, you get $4.8 billion. If it's half that, you get roughly half that.

1.7% vs 5.3%

Assumed annual revenue growth for non-adopters versus practices on paid AI subscriptions. This differential is the entire engine of the $4.8 billion figure.

76%

Practice leaders struggling to find suitable candidates for open roles — up from 71 per cent a year earlier. This one is a survey finding, not an assumption.

The Direction of Causation Is the Whole Question

Set aside for a moment that the modelling comes from a software vendor with an obvious commercial interest — that's worth knowing, but it isn't automatically disqualifying, and vendors often have the best sector data.

The harder issue is that "practices using paid AI subscriptions grow faster" does not establish that the AI caused the growth. Think about what kind of practice buys paid AI subscriptions. It's a practice with budget headroom, a partner who reads about tools, enough process maturity to slot a new one in, and usually enough growth already underway to justify spending. Every one of those characteristics independently predicts faster revenue growth.

The report itself hints at this. It identifies a group of "progressive practices" — described as highly technology-enabled firms already enjoying strong growth and profitability — who communicate with clients more than twice as often as other firms and spend significantly more time developing new relationships. That's a portrait of a well-run, commercially active practice. AI is one of its features. It may not be the cause of its results.

None of which means the effect is zero. It means the honest reading is: firms that operate like this grow faster, and AI is part of how they operate. If you adopt the tools without adopting the operating model — the client contact frequency, the shift from compliance work to advisory — you should not expect the growth differential to follow. Buying the subscription is the cheap part.

The Survey Findings Are More Useful Than the Forecast

The modelling was released alongside MYOB's Accounting Industry Monitor, a survey of 304 owners, directors and senior decision-makers at Australian accounting practices. Several findings there are more actionable than the headline.

Hiring pressure is the standout: 76 per cent of practice leaders reported struggling to find enough suitable candidates, up from 71 per cent a year earlier, and 42 per cent agreed that using technology to free up time for consultancy work could help address the recruitment gap. That reframing matters. The strongest current case for AI in a small finance function isn't margin — it's that the person you were going to hire may not exist at a price you can pay.

On day-to-day benefits, practitioners named faster turnaround (55 per cent), reduced errors (53 per cent) and increased capacity (48 per cent). And 98 per cent said technology-driven automation, process improvement and better communication had contributed to higher revenue or profitability.

That 98 per cent deserves a flag. It's a self-reported perception, not a measured result — and near-unanimous self-reports about a thing people have already invested in are exactly where you'd expect confirmation to show up. Compare it to the "reduced errors" figure: 53 per cent said errors went down. Almost half didn't. Both numbers come from the same survey. The less flattering one is usually the more informative.

What This Means If You Run an In-House Finance Function

Most readers here aren't running a public practice — you're running finance inside an NFP, an NDIS provider or an SME. The translation is direct, and it's about measurement.

MYOB had to assume a growth differential because nobody has a clean before-and-after at sector scale. You don't have that problem. You have one organisation, and you can measure it — if you capture the before-state before you switch anything on. That's the whole discipline, and it's why so many AI business cases end up unfalsifiable.

Pick two or three unit metrics with a denominator. Not "productivity." Things like: hours from period close to reporting pack issued; number of manually-matched transactions per month; percentage of supplier invoices requiring human touch; days from claim submission to payment received. Take a baseline over a full cycle — a quarter, not a fortnight, because month-end is not the same every month. Write the baseline down somewhere that isn't a chat message, with the date.

Then introduce the tool to one process. One. The instinct to roll out across the function is understandable and it destroys your ability to attribute anything. Re-measure after a full cycle, and be prepared for the result to be a wash — that's a valid finding and it's cheaper to discover on one process than five.

The step almost everyone skips is the cost side of that ratio. Track cost per completed task, not cost per token or per seat. An agent that re-reads the same context on every run can be dramatically more expensive per unit of work than the sticker price suggests, and the only way that shows up is if you're dividing by tasks completed.

A note on AI and sensitive data: any pilot involving payroll, participant, client or financial data should start with the vendor's position on training. Confirm in writing whether they train models on customer inputs — and check the AI features already embedded in your existing software, not just the tool you're piloting. The safer default is a plan where customer data isn't retained for training.

The Honest Bottom Line

I don't think the $4.8 billion figure is dishonest. I think it's a reasonable illustration of what compounding does to a growth differential, produced by an organisation that would like you to buy its software, using an assumption it can't yet prove.

The useful version of this story for a finance leader isn't the total. It's three things: the labour market is tight enough that capacity has become the real argument for automation; the practices seeing results are the ones that changed how they work, not just what they bought; and the only ROI number you should trust about your own organisation is the one you measured yourself, against a baseline you wrote down before you started.

If a vendor's forecast makes you want to move faster, that's fine. Just make sure the first thing you do is measure where you are — because in twelve months, the difference between a business case and a story is whether anyone recorded the starting point.

Could you prove what your AI tools have actually saved you?

Setting a baseline, picking metrics that survive scrutiny and reporting them to a board is unglamorous and it's the difference between an investment and an expense. PFL provides senior-level outsourced finance, management reporting, and AI automation for Australian NFP, NDIS, and SME organisations.

Talk to PFL →

Timothy, CPA is Managing Director of Professional Financelink (PFL), providing senior-level outsourced finance, management reporting, and AI automation for Australian NFP, NDIS, and SME organisations. 20+ years in finance leadership across NFP, NDIS and SME.

Sources

Tomorrow: AI is beating humans at more and more cognitive work, and Australian productivity growth has barely moved. Friday's deep dive on what that gap means for how finance leaders should be spending.

Search This Blog

Finance Intelligence