Five Apps, and Nobody Owns the Gaps

1. Five Apps, Four Gaps

Ask an Australian accountant what they run and you will get a ledger and a shrug. Ask what they log into in a week and the number is larger than the shrug suggests. Agile Market Intelligence surveyed 755 Australian accountants and bookkeepers through 2025 and found a median of five platforms per person, with the heavier users on seven or more.

Five platforms is four handoffs. That is the part nobody buys, nobody owns, and nobody prices. Every one of the five was chosen carefully, demoed properly and is good at its job. The gaps between them were chosen by nobody.

Intuit’s 2026 Accountant Technology Survey put the same question to 725 accounting and bookkeeping professionals and found firms averaging ten apps, one in three running eleven or more, and only 41 per cent saying their tools were fully integrated. Two thirds reported feeling overwhelmed by the complexity of their stack at least once a week. That is a US sample, so read the absolute numbers with care, but the shape of it travels, and it matches what you find inside any Australian firm you actually open up.

2. The Layers of a Stack

Stacks look bespoke and are not. Almost every one resolves to the same layers, in the same order, because they follow the path a transaction takes rather than anyone’s preference.

Something happens in the world. A document describes it. Somebody approves it. Money moves. The ledger records it. A report explains it. Every app you own sits at one of those moments, and the useful question about any new one is not whether it is good but which moment it claims.

Read a vendor’s website against that list and the pitch usually resolves in a sentence. Most of the noise in this market is apps that do one layer well advertising as though they do three.

3. The Ledger Is the Decision You Make Once

Every other choice in this paper is reversible in a weekend. This one is not. The ledger decides which apps will connect natively, which accountants will take you on without sighing, and how much of the next five years goes on migrations.

In Australia the practical field is three, and the honest summary is that they are all competent and the ecosystem is the differentiator. Published share estimates vary widely by source, so treat them as direction rather than fact: what is not in dispute is the order.

4. What People Actually Run

The apps below are the ones that keep appearing: in the Xero App Store power lists for 2026, in the Australian usage data, and in the stacks we open up. This is not an endorsement of any of them, and the right answer in each row depends on volume. Read the last column first: buying a layer before you have the volume that justifies it is the most common expensive mistake in the whole exercise.

Two rows deserve a note. In document capture, Hubdoc comes free with a Xero subscription and Dext is the more capable tool, and Australian usage has them close. Practice Protect, which tracked 17.9 million logins across 22,769 Australian accountants for its 2025–26 apps report, had Hubdoc narrowly ahead. Free wins more of these decisions than feature comparisons predict, which is fine right up until volume makes the difference expensive.

In reporting, the same report put Fathom on 36 per cent of the Australian advisory and reporting category, Spotlight on 26, Syft on 20 and Futrli on 18. That is an unusually flat market, which is a useful signal in itself: when four products split a category that evenly, it is because they are close enough that the choice is about how you like to present, not about capability.

5. Two Stacks, Not One

“Accountant stack” means two different things depending on who says it, and conflating them is why so much software advice reads as vaguely wrong. A business runs a stack to keep its own books. A practice runs one to keep everyone else’s, plus a whole layer the business never sees.

Everything in section 4 is the business stack. A firm runs all of it, usually several times over across client files, and then adds this.

6. Where the Value Leaks

Here is the part the app comparison articles never cover, because no vendor sells it. Every app in section 4 does its own job well. The work that is left is the work between them, and it is almost entirely done by a person, in an inbox, at the end of the day.

These are the six joins that cost the most in the stacks we open up. None of them is a missing app. Each is a missing decision, written down and then enforced.

We have written each of these up at length elsewhere: the coding rules in accounts admin, the approval-to-payment queue in payables, the knowing-versus-chasing gap in receivables, the mailbox that absorbs the rest in the accounts inbox, and the job-to-ledger drift in change propagation.

7. How to Choose

Seven rules that have held up across the stacks we have looked at. None of them is about features.

1. 1

   One system of record, defended

   The fastest way to ruin a good stack is to let a second app become the place people look for the truth. Every app either feeds the ledger or reads from it. Nothing gets to be a rival.

2. 2

   Count handoffs, not apps

   Ten apps with nine automatic joins is a better stack than five with four manual ones. The app count is not the metric; the number of places a person retypes something is.

3. 3

   Buy where the queue is

   Walk the layers and find the one with a pile in front of it. That is the layer to spend money on, regardless of which one has the exciting demo.

4. 4

   Interrogate the word integration

   Ask which fields, in which direction, how often, and what happens when it fails. A daily one-way push that silently stops is worse than no integration at all, because you will trust it.

5. 5

   Match the bookkeeper, not the review site

   The app your bookkeeper runs at ten other clients will be configured properly within a week. The better app they have never seen will be half-configured a year from now.

6. 6

   Price the layer, not the licence

   Per-document, per-user and per-entity pricing all look cheap at demo volume. Model it at the volume that made you go looking, then add the hour a week somebody will spend on exceptions.

7. 7

   Every app gets an owner

   A named person who knows the rules inside it and is expected to change them. An app without an owner does not stay configured; it accumulates exceptions until somebody declares it useless.

8. The Order to Build It In

Sequencing matters more than selection, because each layer inherits the quality of the one below it. A reporting tool on badly coded data produces confident, beautiful, wrong charts, and the confidence is the dangerous part.

1. 1Ledger, chosen for the ecosystem and the advisor rather than the feature grid
2. 2Bank feeds, and a chart of accounts somebody actually designed instead of the default nobody has looked at since setup
3. 3Payments in: a pay-now button on every invoice, because it costs almost nothing and moves debtor days immediately
4. 4Payroll, if you have staff. Award interpretation first, everything else second
5. 5Capture, once bill volume justifies it, with the coding rules written down before the trial starts
6. 6Approvals, the moment more than one person can commit money
7. 7Chasing, once reminders have become somebody's mood rather than a process
8. 8Job or inventory system, if the work or the stock lives outside the ledger
9. 9Reporting, last, when the data underneath is trustworthy enough to be worth presenting
10. 10Then go back and automate the joins, which is the only step that compounds

9. What Not to Buy Yet

Stacks get worse in predictable ways. Every item below is something we have watched a sensible business buy for sensible reasons and regret inside a year.

The AI layer deserves its own caution. Adoption is genuinely broad: Intuit’s 2026 survey had 88 per cent of firms using AI for at least one client service, but only around 30 per cent said it was the default in day-to-day work. That gap is the whole story. Most AI in accounting today is somebody trying things, not a process that runs whether or not anyone is watching. The second one is what pays, and it looks much more like plumbing than like a chatbot.

10. Getting Started

1. 1

   Draw it on one page

   Boxes for apps, arrows for data. Do it from memory first, then check. The apps nobody could remember and the arrows nobody could name are both findings.

2. 2

   Mark every arrow that is a person

   Highlighter, one colour. This is the real diagram, and in most businesses it is a surprising amount of highlighter for a stack everyone describes as integrated.

3. 3

   Count the queues, not the hours

   Where does work pile up waiting for somebody? Bills awaiting approval, receipts awaiting coding, invoices awaiting a chase. Queues are easier to measure honestly than time is, and they point at the same layer.

4. 4

   Price the worst join

   Take the single most expensive highlighted arrow and cost it properly: minutes, frequency, error rate, and what an error costs when it escapes. That number decides whether anything here is worth doing.

5. 5

   Fix one join, all the way

   Not one app. One join, end to end, including the exceptions and who owns it. A single join closed properly is worth more than three apps bought at once, and it tells you what the next one will cost.

Figures in this paper come from Agile Market Intelligence’s 2025 Australian accounting technology survey (755 accountants and bookkeepers, weighted by practice size and seniority), Practice Protect’s 2025–26 Australian Cloud Accounting Apps Report (17.9 million logins across 22,769 accountants), Intuit’s 2026 Accountant Technology Survey (725 US professionals) and the Xero App Store’s 2026 Australian power lists. Market share estimates for the ledgers themselves vary widely by source and are treated here as direction rather than fact.