The task
The slow part of a close is rarely the arithmetic. It is finding the document that supports each line and noticing the lines where no document exists. Configure an agent to attach the evidence it can find to each proposed accrual and to say plainly which lines it could not support, so the controller spends the time on judgement rather than on searching.
What you provide
Proposed accrual schedule
The draft schedule for the period, with the cost centre, description, amount and the person who proposed each line.
Supporting documents
Approved orders, delivery confirmations, signed statements of work and correspondence for the period under review.
Your close rules
The cut-off date, the materiality threshold and the evidence your auditors expect for each category of accrual.
Give the agent a clear brief
Use these instructions as a starting point, alongside your approved sources and permissions.
For each line on the schedule, find the documents in the supplied set that support the amount and the period. Quote the document reference and the page. Report a line as unsupported when no document establishes both the amount and the period. Do not adjust amounts, propose new accruals or post entries to a ledger.
How the work comes together
Read the schedule
Take each proposed line with its amount, period and cost centre. Keep the proposer, because an unsupported line usually needs a question rather than a search.
Find the support
Search the supplied documents for evidence of the obligation and of the period it belongs to. An order alone establishes commitment, not delivery, and the two should not be reported as the same thing.
Report what is missing
Return the supported lines with references and the unsupported lines with the reason. The second list is the actual work remaining in the close.
Illustrative example · fictional data
The close list that matters
An illustrative schedule reviewed against a fictional document set, with a cut-off of the last day of the period.
| Accrual line | Evidence found | Status |
|---|---|---|
| Contractor fees, $18,400 | SOW-221 p3 · timesheet approval 28th | Supported |
| Logistics, $6,900 | PO-1190 only, no delivery confirmation | Commitment, delivery unevidenced |
| Marketing, $12,000 | No document in the supplied set | Question for the proposer |
The second line is the interesting one. An order proves the company committed to spend, not that the service arrived inside the period.
The human review
A controller decides whether each line is accrued, adjusted or removed. The agent supplies the evidence and the gaps; the accounting judgement and the ledger posting remain yours.
Set it up for your team
Define supporting evidence
Write down what counts as support for each category of accrual, and where an order, a delivery confirmation or a signed statement of work is sufficient on its own.
Scope the document set
Give the agent the period’s documents and the immediately surrounding weeks. Cut-off errors hide just outside the period you are closing.
Agree the threshold
Set the materiality below which a line is listed but not investigated, so the output stays the size of an actual close checklist.
Check the result before scaling
Run the playbook against a period you have already closed and audited. Compare the evidence it found with the evidence in the file, and count the unsupported lines it missed before using it in a live close.
Keep the data in the right place
Close working papers reveal unreported results, commitments and contract values ahead of any announcement. A fully local deployment keeps them inside your network for the period when they are most sensitive.
Explore deployment optionsQuestions & answers
Does this replace our close checklist?
No. It gathers evidence against a schedule you already produce and shows the gaps. The checklist, the judgement and the sign-off stay with your close team.
Can it propose accruals we forgot?
Not reliably. It reviews the lines on the schedule you supply. Finding an obligation nobody recorded is a different problem and needs a different, carefully evaluated configuration.
Is the output suitable for auditors?
It produces references to your own documents, which is a useful starting point. Whether that satisfies your auditors is a question for them, and every reference should be checked before it is relied on.