Workpapers are the schedules, reconciliations and source documents that show how each balance in a set of year-end accounts was reached. A complete file lets a reviewer who has never met the client agree every material balance to evidence without asking the preparer a single question.
That second sentence is the standard, and most files miss it. The numbers are usually right. What is missing is the proof, and the reviewer pays for the gap in hours: either they re-perform the work, or they sign on trust and carry the risk.
What workpapers are for
The trial balance says what the ledger holds. The financial statements and the tax return say what the firm is prepared to put its name to. Workpapers are everything in between: the record of how the firm got from one to the other, and why it believes the result.
Three people rely on that record. The reviewer relies on it this year, because under the Tax Agent Services Act supervision and reasonable care sit with the registered agent, and a signature is a statement that the work was checked. Next year’s preparer relies on it to understand what was done and why. And whoever asks later relies on it too, whether that is the client, a new adviser or a regulator, because by then the file is the only witness left.
What a complete company file contains
The exact set depends on the entity, and a good preparer selects workpapers for the accounts this client actually has, not the ones last year’s file happened to carry. For a typical company job the file holds:
- An index and lead schedule that ties every workpaper back to the trial balance, so nothing in the ledger is left unexplained.
- Bank reconciliations agreed to the bank statement at balance date, with reconciling items listed and followed up.
- Receivables and payables listings agreed to the ledger, aged, with old or unusual items explained.
- A fixed asset register and depreciation schedule, with additions traced to invoices and disposals accounted for.
- Loans and borrowings agreed to lender statements, with interest reproved and the current and non-current split supported.
- Payroll liabilities: PAYG withholding and superannuation reconciled to payroll reports and to what was actually paid.
- A GST reconciliation between the activity statements lodged for the year and the ledger.
- Related-party and director loan accounts, including Division 7A loan balances, agreements and minimum yearly repayments.
- An income tax reconciliation from accounting profit to taxable income, and the franking account.
- Equity and dividends, agreed to minutes and to the company’s records.
- Journals, each with its reason and its support attached.
- Queries and answers: what was asked of the client, what came back, and what is still open.
A file can hold every one of those and still fail, because presence is not the test.
The three tests a reviewer applies
Does it agree? The total on the workpaper matches the trial balance. This is the test every file passes, and it proves the least, because a schedule built from the ledger will always agree to the ledger.
Is it proved? The balance is supported by something the ledger did not produce: a bank statement, a lender’s letter, a signed agreement, an invoice. A loan account that agrees to the ledger tells you the bookkeeping is consistent. A loan account that agrees to the lender’s statement tells you the balance is right.
Can a stranger follow it? Hand the file to someone who has never seen the client. If they need the preparer in the room to understand a number, the workpaper is unfinished. The explanation that lives in someone’s head leaves when they do.
Where files go wrong
The same faults turn up in firm after firm, and none of them is about competence.
Last year’s file gets rolled forward and this year’s numbers dropped in, so the workpapers describe the client as it was. A small reconciliation difference gets plugged to a suspense or sundry account because the job is already over budget. The evidence exists but lives in an inbox, so the file points at a document nobody else can open. Client queries are answered by phone and never written down. And where support is missing, the gap is quietly filled with an assumption that is not recorded anywhere.
The last one does the most damage. A reviewer can deal with a stated gap: it becomes a query, a note, or a judgement call they knowingly make. A hidden gap gets signed.
Where AI fits
Preparing workpapers is the part of a company job that software can now take on, provided the output meets the same three tests. That is how we built MindLedger’s Workpaper Processing module, which is live for company jobs.
The agents extract the ledger and source data, select the workpapers for the entity, prove the balances and work the exceptions. What comes back is a work packet: workpapers, draft financial statements, a draft company tax return for your review, and an assumptions log, with every figure traced to its source. Open gaps are listed, not filled.
None of it changes who is responsible. Under the TPB’s AI guidance, TPB(GS) 55/2026, the practitioner stays responsible for AI-assisted work and must review it before relying on it. Your accountant reviews and signs off every job, and the draft return is lodged in your own software. We have written before about why that review only works when the workings arrive with the answer, and about how a job runs from handover to packet.
Pull the last company file your firm signed and hand it to someone who has never seen the client. How far do they get before they need to ask a question?