How This Works in a Real Practice
Three illustrative engagements showing the shape of the problem we solve, the approach we take, and where we deliberately stop.
These are illustrative scenarios, not client case studies. They are composite examples of the kind of practice this service is built for. They describe no named client and contain no measured results. Where you would normally expect a before-and-after number, we have left it out on purpose — see the note at the bottom of this page.
A 12-person bookkeeping practice
The Situation
Supplier invoices arrive at the practice from every direction — client inboxes forwarded on, supplier portals, photographed dockets, and the occasional envelope. Four staff spend most of their week keying those invoices into client ledgers and coding them. The work is necessary, it is charged at a rate the clients push back on every year, and it is the reason the practice has never had the capacity to start the advisory service the partners keep talking about.
The Approach
- Connect each client ledger in read-only mode and let the AI learn how the practice has historically coded each supplier
- Point the document flow at the AI — forwarded emails, portal collection, and the existing capture tool all feed the same queue
- Turn on auto-posting first for recurring suppliers under a modest dollar threshold, keeping everything else in review
- Widen the threshold client by client as the exception rate on each one settles
What Changes
The processing work shifts from keying to reviewing. Staff clear a queue of confident suggestions and spend their attention on the flagged exceptions — new suppliers, unusual GST treatment, amounts outside the pattern. The capacity that frees up is what the practice uses to start the advisory work, rather than to take on more compliance volume.
What stays with the humans
Every ledger the AI writes to is still reconciled and reviewed by the practice. Nothing is lodged.
A four-partner suburban accounting firm
The Situation
The quarter runs the firm rather than the other way around. In the fortnight before a BAS deadline, everyone drops advisory work to chase source documents, resolve coding queries, and reconcile ledgers that have drifted since the last quarter. Overtime is budgeted for. The partners know exactly which clients will be the problem and know they will find out too late anyway.
The Approach
- Run document chasing continuously through the quarter instead of in a two-week panic before the deadline
- Flag inconsistent GST treatment across similar transactions as it occurs, not at review time
- Surface unreconciled items and document gaps on a live view, per client, so problem files are visible in week two rather than week eleven
- Assemble the working papers and document trail for each client into one reviewable file ahead of the BAS agent picking it up
What Changes
The work that used to compress into a fortnight spreads across the quarter, and the BAS agent starts their review from a prepared file rather than a scavenger hunt. The deadline stops being an event. The firm gets its advisory fortnight back four times a year.
What stays with the humans
The BAS agent reviews, applies professional judgement, signs off, and lodges. The AI never touches the ATO.
A solo bookkeeper with about 30 clients
The Situation
The bottleneck is not the bookkeeping — it is the chasing. Missing receipts, unanswered coding questions, statements that never arrive. Each individual follow-up takes two minutes; thirty clients of them takes a day a week. The alternative is doing the work with gaps in it and fixing it later, which costs more.
The Approach
- Automate the follow-up cycle: missing documents and open coding queries chased on a schedule, in plain English rather than accounting jargon
- File replies and attachments back against the right client and period automatically
- Escalate to a personal call only when a client goes quiet or the answer is genuinely ambiguous
- Keep coding suggestions in review-only mode indefinitely — a solo operator reviewing a queue is faster than one keying it, without giving up control
What Changes
The chasing runs without the bookkeeper doing it, and the personal contact gets spent where it actually matters. Practically, this is the scenario where automation substitutes for a hire the practice was never going to make.
What stays with the humans
Every coding decision is reviewed and posted by the bookkeeper. The AI drafts the chase; it does not own the client relationship.
The Engagement Model
How an Engagement Actually Runs
Whatever the scenario, the sequence is the same. You never hand over a client ledger on the strength of a demo.
Scoping Call
We map the admin your practice carries today and are direct about which parts AI should take and which it should not. If automation is not the answer for your book, we will say so.
Read-Only Pilot
The AI connects to one or two client ledgers and proposes without writing. You watch it for a few weeks and form your own view of the suggestion quality on your data.
Guardrails & Go-Live
You set the dollar thresholds, new-supplier rules, and per-client overrides. Auto-posting starts narrow — usually recurring suppliers under a modest limit.
Progressive Expansion
You widen the remit as the exception rate on each category earns it. Client by client, threshold by threshold. Never a switch you flip and hope.
Why There Are No Numbers on This Page
Software marketing in this category runs on invented statistics. Hours saved, percentages reclaimed, accuracy rates that no one could reproduce, a testimonial from a firm that does not exist. You are accountants. You know what an unsourced number is worth.
So we do not publish results we cannot substantiate. What automation is worth to your practice depends on things we cannot know from a web page: your client mix, the state of your ledgers, how much of your document flow is already digital, and how much you actually choose to automate once you see it working. A firm with clean Xero files and a captured document flow gets a very different answer from one still receiving photographed dockets.
What we will do is sit down with your practice, look at your real volumes, and give you an honest estimate you can check against your own timesheets. If we ever publish a client result on this page, it will be a real firm, named, with their approval, and with the method behind the number stated plainly. Until then, this page stays illustrative.
Does One of These Sound Like Your Firm?
Book a scoping call. We will map the admin you are carrying and tell you which parts are worth automating — including if the answer is not many.