Blog

Div 7A Loan Tracking: An AI-Assisted Register

July 2026 · 7 min read · Industry Guide

An open ledger book illustration with one loan repayment row highlighted in terracotta, a pen resting across the page.
← Back to all posts

Most private company loans start the same way. A director needs cash, the company has it, and someone moves the money without a second thought about Division 7A of the tax law. By the time a bookkeeper in Sydney or Melbourne raises it at tax time, the loan has usually been sitting untracked for a year or more, and the minimum yearly repayment has already been missed.

That missed repayment isn't a paperwork problem. Under Div 7A rules, an underpaid loan is treated as an unfranked dividend to the shareholder, taxed at their marginal rate with no franking credits to soften it. For a $180,000 loan, a single missed year can turn a manageable repayment into a tax bill the client never budgeted for.

Why Div 7A tracking breaks down in spreadsheets

Every Australian accounting firm has a Div 7A spreadsheet somewhere. The problem is rarely the template. It's keeping it current across dozens of clients, each with their own loan agreements, start dates and repayment history.

  • The benchmark interest rate changes every year. If last year's rate is still sitting in the formula, every minimum repayment calculation downstream is wrong.

  • Loan agreements live in email, not the ledger. The 7-year unsecured term (or 25-year secured term) is easy to lose track of when the agreement is a PDF from three years ago.

  • Part-year loans and top-up advances complicate the minimum repayment formula, and a manual spreadsheet rarely handles both correctly.

  • Nobody notices an underpayment until the 30 June review, by which point the deemed dividend has already crystallised.

Div 7A loans show up in more ways than a straight director drawdown. Unpaid present entitlements from a trust to a corporate beneficiary, private use of a company asset, and informal top-up advances mid-year all fall under the same rules, and each one needs its own start date and repayment schedule in the register. A firm running twenty corporate clients might have sixty or seventy of these loans active at once, and a handful of missed rows is where most of the risk sits.

What an AI-assisted register actually looks like

Firms using Claude for this aren't asking it to give tax advice. They're using it to do the thing spreadsheets are bad at: keeping a register current, flagging exceptions, and producing a clean summary a director can actually read.

A typical setup connects Claude to the practice's Xero or client ledger data, plus a small reference table of the ATO's published benchmark interest rate for each income year. From there, Claude maintains one row per loan with the opening balance, the applicable rate, the minimum yearly repayment, and what's actually been repaid so far.

The three numbers Claude needs to track per loan

  • Opening balance at the start of the lodgment year, carried forward automatically from the prior year's closing balance.

  • The benchmark interest rate for that income year, pulled from a maintained reference rather than typed in fresh each time.

  • Cumulative repayments made against the minimum yearly repayment, updated as bank data or journal entries come through.

Setting it up with Claude

This isn't a large build. A practice manager can set up a working version with Claude Cowork or Claude Code in an afternoon, working from three inputs.

  • Client loan register. A simple table with client name, loan start date, original amount, and agreement term, pulled from existing files or a Google Sheet.

  • Repayment source. Bank feed data or Xero journal lines that show what's actually moved back to the company against each loan.

  • Benchmark rate table. Updated once a year when the ATO publishes the new rate, rather than buried in a formula nobody remembers to check.

From there, Claude can generate a per-client summary showing the minimum repayment required, what's been paid, and the shortfall if any, with plain-English notes a client-facing accountant can drop straight into an email. A Brisbane bookkeeping firm running this across 40 corporate clients turned a two-week annual scramble into a standing register they check monthly.

What this looks like at 30 June

The value isn't the register itself, it's catching the shortfall in March instead of discovering it in July. When Claude flags that a $45,000 loan is $6,200 short of its minimum repayment with three months of the income year left, there's still time to top up the repayment before it converts to a deemed dividend.

  • A one-page loan summary per client, ready to attach to a review email.

  • An exceptions list showing which loans are behind on minimum repayments, sorted by shortfall size.

  • A reminder trigger set for well before 30 June, not the week of it.

None of this replaces the accountant's judgment on how a loan should be structured or whether a complying agreement needs updating. What it removes is the manual chase: the register stays accurate on its own, and the exceptions surface early enough to actually do something about them. If your firm is still tracking Div 7A loans in a spreadsheet that only gets opened once a year, that's a reasonable place to start automating.

If you want a hand setting this up for your practice, book a brainstorm session and we'll walk through what a working register would look like with your existing client files.

Ready to move from AI pilot to production?

We help mid-market Australian businesses deploy AI automations that actually reach production and deliver measurable ROI.