The Quiet Reinvention of the Small Accounting Firm

It's a Wednesday in Parramatta. The senior partner of a two-person accounting firm — the kind with a fluorescent-lit reception desk and a kettle that's older than the GST — is finishing a licence audit for a client. A 240-seat retailer on Dynamics 365. Mostly E3 seats. Mostly underused. The audit took her eleven working days. She billed $18,400. The fee was fair. The work was honest. The client paid without complaint.

What she doesn't yet know is that an agent built around the same Microsoft entitlement data, the same client telemetry, the same Australian Tax Office rulings, can do the recommendation pass in eleven minutes — and produce a sharper, better-defended report than the one she just delivered. Not because it is smarter than her. Because it is tireless, and because it never forgets a single audit it has ever read.

This is not a story about the small accounting firm disappearing. It's a story about the small accounting firm — and what it has always been — finally getting the leverage it was promised twenty years ago when the cloud was supposed to fix everything and didn't.

The state of the frontier

Three things are converging in the back rooms of Australia's roughly 30,000 small accounting and advisory firms right now.

First, the licence and subscription estate of their average client has become genuinely unmanageable. A 200-person business now runs somewhere between forty and eighty SaaS contracts. Microsoft alone offers more than a hundred distinct Dynamics 365 and Power Platform SKUs, each with a different per-user economic envelope. The CFO can no longer answer "how much are we paying per active user" without an entire week of someone's life. Globally, analyst estimates put unused SaaS spend somewhere between 25% and 35% of total subscription outlay. In Australia that is, conservatively, billions of dollars a year sitting in licences nobody opens.

Second, the buyer is asking different questions. The CFO at that 240-seat retailer doesn't want a quarterly licence audit anymore. She wants the licence count to be continuously correct. She wants the advisor to notice the minute Sarah from marketing is provisioned an F5 seat she'll never use. She wants the recommendation, not the report.

Third, the toolkit has changed. The same private-trace, learning-loop pattern that has now displaced 74% of failed enterprise agent deployments with the 26% that survive is sitting one repository away from any small firm with the curiosity to use it. Entitlement graphs, usage telemetry, ruling libraries, partner-program rebates — all of it can be wired into a single loop that holds this firm's judgement, not a generalist model's.

The bottleneck used to be the audit. The bottleneck has become the advisory. And advisory is, definitionally, what the small accounting firm was always for.

The compound mechanism

Here is what changes when one agentic foundation sits under the practice rather than under a single product.

The licence audit becomes a continuous service, not an annual event. The advisory conversation becomes monthly, not quarterly. The proposal generation, the rebate optimisation, the security-posture flag, the unused-seat reclamation, the executive briefing — all of these used to require either separate specialists or separate software. Now they sit on one shared loop. The marginal cost of the next service line collapses. The marginal cost of the next client collapses. The marginal cost of the next jurisdiction collapses.

That is what the compound model means in practice for a two-partner firm. The senior partner is not "doing audits with AI." She is running a small house of services — audit, advisory, procurement, briefing, post-implementation review — on one foundation. The same loop that watches one client's Dynamics estate also watches the firm's own books. The same loop that drafts the recommendation also drafts the client podcast that explains it. One team. Many products. Three uncorrelated markets within professional services itself.

This is not a SaaS pitch. It is a re-architecture of how a firm of this size makes a living. Twenty years of cloud transformation did not, in the end, change the unit economics of a small advisory practice in any meaningful way. The compound model can — because for the first time, the marginal cost of doing the next kind of work for the same client falls to near zero.

Who benefits

Name them, because the line is what matters.

The two-partner firm in Parramatta that has been quietly losing tenders to one of the Big Four for five years. The firm now wins the bake-off because its continuous licence-correctness service costs the client one-third of an annual audit and produces better numbers.

The CFO at the 200-person retailer who finally has a number she trusts on slide three of the board pack — and can defend it because the trace is auditable, the ruling library is current, and the agent's recommendation is reproducible at any point in the year.

The IT director who has been losing $400,000 a year to unused F5 and E5 seats without ever quite being able to prove it to procurement. She can prove it now. With names attached. By Friday.

The founder of a 25-person logistics business in western Sydney who could never justify a CFO and never trusted a bookkeeping platform on its own. She now has both — paid by the hour, attentive, accountable — wrapped around her own books.

This is the livelihood payoff that matters. The small firm doesn't shrink. The advisory profession doesn't hollow out. The mid-market client gets the operational hygiene that until last year was the exclusive privilege of the ASX 50.

This is what good looks like, in this lens. Not a frontier model demo. A Parramatta bookkeeper with better margins, a Penrith CFO with fewer surprises, a Wollongong founder with a back office she can actually run.

What stays human

The recommendation is automatable. The relationship is not. The audit is automatable. The judgement about which line item to bring up at the partner dinner is not. The reconciliation is automatable. The decision to call the client at 7pm because something doesn't feel right is not.

Three things stay human in this picture, and they are the three things that have always been the small firm's actual moat. The first is taste — the partner's read on which finding to lead the conversation with, and which to bury on page seven. The second is trust — the fact that the CFO returns this partner's calls because of twelve years of being right when it mattered. The third is accountability — somebody whose name is on the engagement letter, who carries the consequence if the advice is wrong.

No agent absorbs any of these. They are the part of the work that pays. They are the part of the firm that compounds across decades, not quarters. Anybody who tells you the agentic era is going to flatten advisory has not spent a Wednesday afternoon in a small accounting practice trying to talk a founder down off a poor decision.

What the agent does is finally give that judgement somewhere honest to land. The partner spends her afternoon on the conversation, not on the workbook.

What we're building toward

Inside Agenticality, this is exactly the bet MetaScan-D365 is built around. Not "AI for licence optimisation." A continuous, private-trace, learning-loop service that any advisory practice can run as their own — with their judgement in the seat, their client roster on the shelf, and the boring six-figure-a-year savings landing in the client's account month after month.

iFQ sits one floor up — the same compound foundation, redirected at the household. Three currencies, two earners, one mortgage, four super accounts, a small business in the spare room. The same private-trace, the same loop-as-IP discipline. The same refusal to use anybody's data to train anybody's model.

PodGenCast sits one floor down — because the briefing, the client recap, the partner's monthly five-minute update, can all be heard rather than read. The player under this essay is the product running on itself.

Three uncorrelated markets — small advisory, household financial sovereignty, audio briefings. One correlated capability. Diversification without dilution.

The Parramatta partner doesn't need any of this to be a paradigm. She needs it to be Wednesday-evening practical. So that is what we are building. Live. Continuously improving. With her, eventually — not at her.

— Agenticality · 21 June 2026 · Sydney