> ## Documentation Index
> Fetch the complete documentation index at: https://control-dev.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.
# Review the monthly close with an AI client
> Watch one ask during the close become a full ledger sweep: wrong-account bookings and missing accruals found, tabled with fixes, and correcting entries offered on approval.
One ask during the close checklist — "are there missing accruals or transactions in wrong accounts?" — and the
client sweeps the trailing twelve months of the group ledger, checking every booking against the account
definitions.
The recording has narration, and the figures, entities, and account names in it are illustrative. Nothing in the
video is a target or a benchmark.
## What happens in the recording
1. The user asks, mid-checklist, for missing accruals and transactions in wrong accounts. The client loads a
ledger-review skill and gets to work.
2. It sweeps all 8,141 group transactions from the trailing twelve months, checking each against the group account
definitions — with stated audit recipes, like verifying one-hit insurance premiums against prepaid schedules and
confirming seasonal pay patterns before excluding them.
3. The verdict: 21 findings — wrong-account bookings that misclassify the P\&L, and unexpired costs expensed up
front that belong on accrual schedules.
4. Every finding is tabled: date, entity, counterparty, where it was booked, the suggested fix, the amount — plus
the evidence, links to the transaction views in Control, and what was reviewed and deliberately *not* flagged.
5. The client states the profit impact after correction and offers to draft the correcting entries as proposals for
approval.
## Try it in your workspace
1. [Connect an AI client with MCP](/product-docs/ai/mcp-server) — read access finds the issues; write access lets
it propose the corrections.
2. Ask the close question directly: "any missing accruals or bookings in wrong accounts this period?"
3. Require evidence per finding — which account definition is breached, and a link to the transactions.
4. Have corrections drafted as proposals and approve them yourself.
## A tailored skill makes the difference
The review is only as good as the client's idea of what belongs in each account. A `SKILL.md` written for your
process — your entity account list, what each account should and should not contain, known seasonal patterns that
must not be flagged, materiality thresholds — turns a generic sweep into your review. In the recording that is why
the client can cite a breached account definition per finding and correctly leave seasonal holiday-pay spikes
alone. Keep a file this specific local to the workspace it describes — shared skill files stay generic, per
[Use skills with Control MCP](/product-docs/ai/mcp-skills).
Example prompts: "Go through this month's close: missing accruals, wrong accounts, anything expensed that should
be spread." "For every finding, cite the account definition and link the transactions." "Draft the correcting
entries as proposals — do not book anything."
## Troubleshooting
### Seasonal or one-off items keep getting flagged
Document the known patterns — holiday pay, bonus rounds, annual premiums — in your skill file, with the months they
hit. The client should then confirm the pattern and move on, as it does in the recording.
### Findings arrive without evidence
Ask for the rule behind each finding — which account definition it breaches — and a transaction-view link. A
finding you cannot inspect is not actionable during a close.
## Next steps
Write the skill file that encodes your close process.
Turn the clean close into a budget-vs-actuals package.
Report on the period once the ledger is clean.
Investigate the movements the corrections change.