Journal entries screened for fraud and errors
Every approved journal is checked for self-approval, duplicates, dormant accounts and odd amounts, then explained plainly.
Live in partWhat does Anomaly rules engine + LLM narrator do? Spot possible fraud and errors soon after approval, not at audit. Each flag points to one voucher, not a haystack.
How it works
Once a journal is approved, six fixed rules check for self-approval, possible duplicates, dormant accounts, unusually large amounts, postings at period end and weekend postings. Anything found goes to a list of exceptions and to your inbox, with the voucher number. Only when a rule flags something does AI write a plain-language explanation.
It also records self-approvals below the ₹5 lakh limit for needing two people, which the approval rules deliberately allow.
Step by step
- An accountant posts a journal. It goes for approval in the ordinary way. Nothing about the approval changes.
- A second person approves it. The journal is then in the books.
- Six fixed checks run straight after approval. They check for the same person posting and approving, a possible duplicate (same amount, same party and date within seven days), a dormant account (nothing in 90 days), an amount more than two standard deviations from that account's 30-day average, an entry within a day of month-end, and a weekend posting.
- Any finding is saved against the voucher number. It goes to a list of control exceptions and to the AI inbox, and into the daily digest. If the checks fail to run, the journal stays approved.
- Only then does AI write a plain explanation. If nothing was flagged, no AI is used. When something was flagged, the AI explains in a sentence why it is worth a look, for example that the amount matches an invoice cleared three days earlier.
- The finance head follows up. They open the journal from the list and ask the preparer. The finding shows as open until it is cleared.
See the six checks run on a sample of your approved journals.
What you see on screen
The list is called Control exceptions, in Finance.
| Screen or button | What it shows or does |
|---|---|
| Control exceptions list | Newest first. Columns: Caught, Rule, Severity, Journal (the voucher number, which opens the journal), Finding and State (open or cleared). |
| Filters | By rule, by severity and by open or cleared. |
| Rule names | Shown as they are, such as SOD_VIOLATION and POSSIBLE_DUPLICATE, so an auditor and the system use the same word. |
| Empty list | Means nothing matched. A failure to load is shown as a message, not as an empty list. |
Who uses it
| Who | When |
|---|---|
| Finance head | Reads the exceptions weekly or after month-end. |
| Internal auditor or your CA | Uses the list as a starting point for sampling. |
| Owner | Sees the self-approval cases, which matter most. |
Related
- Finance and tax: journals, approvals and the books.
- Financial statements that check themselves: the checks on the statement once it is drawn.
- Changes recorded: who, what and why: the trail behind a flagged record.
Questions
What is Anomaly rules engine + LLM narrator in Kinnow?
Every approved journal is checked for self-approval, duplicates, dormant accounts and odd amounts, then explained plainly.
What changes for the business?
Spot possible fraud and errors soon after approval, not at audit. Each flag points to one voucher, not a haystack.
Is it available today?
It is live in part: the core is built and parts are completing with early customers.
Does Kinnow block a journal that looks suspicious?
No. The checks run after approval and tell you. The approval rules themselves are what stop a posting, and the flag shows you where those rules allowed something worth reviewing.
Why is self-approval flagged if my rules allow it?
Your approval limits may allow one person to post and approve a small entry. Kinnow still records it, so you can see how often it happens and decide whether the limit is right.
More in the Answers in seconds
Ask your business anything
One search box on every screen finds records instantly and answers questions from data you're allowed to see.
Answers in seconds · Code on Wages restructuring agentSee who fails the new Labour Codes
See which salary structures break India's new Labour Codes and what fixing each one will cost.
Answers in seconds · Asset operating log + Operating Efficiency WatchCatch fuel theft and failing generators
Daily generator logs reconcile fuel to the tank reading and flag a machine whose output per litre is falling.
Answers in seconds · Clause extraction → Obligation ledgerContract promises tracked before they're broken
Kinnow reads your contracts, suggests key clauses, and tracks every promise, from service levels to reports, before breach.
Answers in seconds · Reconcile against your documentsCheck a spreadsheet against your signed paperwork
Upload a spreadsheet; Kinnow checks every row against your documents and lists what disagrees.
Answers in seconds · Vault (Signals → Vault → Compose)One secure home for company documents Live in part
Every company gets a controlled document store: no duplicates, every access logged, shares you can revoke.
Tell Kinnow what slows you down.
Answer three questions. Get the applications, the agents and a two-week plan for your business.