AI for bookkeepers. The chasing, done without you.

It does the chasing. The documents before a BAS, the unanswered queries sitting in a reconciliation, the onboarding questions asked of every new client, and the monthly update that tells a client where things stand.

AI for bookkeepers is that chasing, handled, rather than anything that touches the reconciliation itself.

Two different things get called AI, and only one of them matters here. The first is automation, which means a job that starts on its own: a period closes, so the document list goes out. The second reads what a client wrote back and answers it in your words, which is what turns a reminder into something people actually reply to.

What it takes off you first

• The documents you need before every BAS.

  The list goes out on the same day each period, and the reminders keep going until the last receipt is in.

• The reconciliation queries nobody answered.

  The unexplained transactions go to the client as a plain list, with a reminder on each one nobody has answered.

• The onboarding questions for a new client.

  Structure, software, bank feeds, who approves what. Asked and checked before the first period rather than over a fortnight of emails.

• The monthly update every client expects.

  What was reconciled, what is outstanding and what is owed, sent without anybody assembling it by hand.

• The payroll and tax questions asked every cycle.

  Timesheets that have not come in, super deadlines, the same tax questions asked before every lodgement. Chased on a schedule rather than remembered.

All of that is work your practice already does. It happens on the day it should, instead of in the week somebody finally gets to it.

They go into chasing, and they cluster. A bookkeeping practice loses more time to following up documents and queries than to any other unbillable job, and it loses the most of it in the fortnight before a lodgement date.

Watch what happens to a document request across a client list. It goes out on Monday. Some clients send everything, some send half, and some send nothing at all. Working out which is which, and then writing to each of them separately, is the job that eats the fortnight.

In a quiet month that is an hour. Across a full client list in the run up to a BAS it is the reason the week runs out, and it grows at exactly the same rate as the billable work does.

What that costs you

• Hours you cannot put on an invoice.

  Every hour on chasing is an hour not on a file, out of the same working week.

• Work that arrives at the deadline.

  A file that could have closed early gets closed at the deadline, so the same clients occupy a longer period for the same fee.

• The person you are paying to do the chasing.

  The chasing lands on whoever is easiest to interrupt, which is rarely the best use of them.

The Institute of Certified Bookkeepers sets out what a practice is expected to hold and when in its guidance for Australian bookkeepers, and the practices absorbing that without adding staff are the ones that put structure around the chasing.

Nothing here asks you to change systems. The work happens in the tools your practice already pays for, and the value is in the handoffs between them rather than inside any one of them.

Xero, MYOB or QuickBooks holds the ledger. Dext or Hubdoc does the data capture. Something else again holds practice management and payroll. Each of those accounting tools does its own job properly. What none of them does is move information from one to the next, which is why a person is still doing that by hand.

Where the handoffs are

• Between the ledger and the client.

  An unexplained transaction sits in the reconciliation until somebody writes to the client about it, one line at a time.

• Between capture and review.

  Receipts get read automatically, and somebody still checks the ones the software could not.

• Between the bank feed and the file.

  A feed drops out or a rule stops matching, and nobody finds out until the reconciliation will not balance.

• Between the engagement and the work.

  The letter goes out, and somebody checks who has signed before anything can start.

Individually each of those takes a couple of minutes. Across a full client list through a BAS period they are most of the reason the fortnight runs out. The audit maps those accounting workflows in your own stack and puts a number beside each one before anything is built.

None of it depends on which ledger you run. A bank reconciliation query leaves Xero the same way it leaves MYOB, and the automated bookkeeping around it reads the same data either way. Practices on two different systems get the same build, pointed at different software.

It does not do the bookkeeping. It makes no judgement call you would have to stand behind, it holds none of your registration, and it decides nothing about how a transaction should be treated.

This matters more here than in most industries, because the work goes out under your name and your BAS agent registration. The automation handles the layer around it and stops at the edge of anything requiring your opinion.

Where it stops, deliberately

• Anything needing your judgement.

  The item is flagged and routed to you rather than answered. A confident wrong answer is far worse than no answer.

• Anything you lodge.

  It will chase and it will assemble. It will not decide what goes on a statement, and it will not lodge one.

• Client data going where it should not.

  Financial data is not handed to a public AI tool. The builds run on your own accounts with the handling agreed in writing first.

Where the work is genuinely repetitive and written down, it runs. Where it is not, it stops and waits for a person, and every run is logged so you can read what it did.

The chasing that only happens when somebody remembers goes on a list, and the job costing the most time comes off it first. The wording gets approved before anything is built, and it runs watched before it runs alone.

The first one is the document chase, because it is the one that costs the most in the fortnight it matters. The next starts once that one is running.

1. The chasing gets written down properly

Half a morning. What goes out, when, to whom, and what happens when a client sends half of it. Most practices have never had this written down anywhere.

2. The job costing you the most time goes first

A rough number goes beside each one. The pre-BAS document chase usually wins on its own, and it wins by a distance.

3. You approve the wording before anybody builds anything

Every message goes out in your practice's name, so it gets written the way you would write it and signed off before it exists as anything but words on a page.

4. It connects to what you already run

Your ledger and your practice management can already tell another system when something happens. Nothing gets replaced and nothing gets migrated.

5. It runs watched before it runs alone

For a fortnight, with a person approving each message before it sends. That fortnight finds the client situations nobody predicted, before a client meets one.

6. The next one starts once the first is boring

A job is finished when nobody in the practice is thinking about it any more. That is the signal to start the next one, rather than a date on a plan.

Writing down how the chasing actually happens takes longer than building it. The first job runs watched inside a fortnight, and each one after it is shorter because the connections already exist.

The order it happens in

• The writing down comes first.

  How a request actually goes out today, including the parts nobody has put on paper. This is the slow half.

• Then it gets built and watched.

  It runs with a person approving each message, long enough to meet the situations nobody predicted.

• Then it runs on its own.

  You hear about it monthly: what ran, what it replied to, and anything that failed and stopped.

The first thing a practice notices is that the chasing stops being anybody's job. What follows that is a BAS period where the documents arrive earlier, because the reminders went out whether or not anyone had time to send them.

Support afterwards is part of the fee rather than a separate line. When a bank feed drops or a client changes accounting software, somebody who knows the build fixes it, and the monthly note tells you it happened. The same note carries the only insights worth having about the chasing: how many clients answered first time, and which ones still needed a person.

Ask where it runs, whose accounts it uses, what happens to client data, and what you keep if you stop paying. Those four answers separate a build you own from a subscription you rent.

These are worth asking of anybody, including this one, so the answers given here sit under each question.

• Where does it actually run?

  On your own accounts and a server you can point at, rather than inside a platform you rent. The first one is yours and the second one is not.

• What happens to client financial data?

  It stays inside systems you control, and it is not handed to a public AI tool. Agree the handling in writing before any money changes hands.

• What happens when it breaks?

  A bank feed drops or a rule changes and a job stops. Ask who finds out first, how fast, and whether a half finished message can reach a client.

• What do I keep if I stop paying?

  A build keeps running afterwards. A subscription stops on the day the paying stops, and you start again from nothing.

• Am I locked into a contract?

  Not here, it runs month to month. Any arrangement you cannot leave at the end of a month is worth a second look.

Pricing is worth asking about in the same breath. One monthly fee that covers the building, the support and the monitoring is a different thing from a per seat licence on a piece of bookkeeping software, and the second one grows every time your practice does.

All of this can also be built in house, and plenty of practices do it well. The six steps above are the same either way.