Can the AI be trusted? Here is the number.
Every month, a report card for your books, computed from the records DayZero already keeps and shareable with one link. Nothing in it is typed by hand.
- Who coded itRules · agents · peopleEvery transaction in the month, by who categorized it.
- What was revertedReversal rateRows undone, divided by rows automation changed.
- How fast it closedBusiness daysFrom period end to the close, against your target.
- The gradeA to DOne published rubric. The same numbers always get the same grade.
The books, as a report card.
Six figures and a grade. Each card names the record it was computed from, and a figure the books do not track yet says so instead of showing a zero.
Auto-coded
96.2%
Reversal rate
0.17%
Closed in
2 business days
| Who coded it | Transactions | Share |
|---|---|---|
| Bank rules | 1,104 | 59.9% |
| Agents | 668 | 36.3% |
| People | 70 | 3.8% |
| Approvals | 41 approved | 2 rejected |
Facts from the ledger, not claims from a brochure.
The scorecard reads the same tables the audit trail does. There is no field a person can edit.
Who coded it
Every transaction records whether a bank rule, an agent, a payment match or a person set its category. The split and the uncategorized remainder come straight from that record.
What automation did
Every unattended run is logged with the rows it changed. Undo a run and the reverted rows count against the month. The reversal rate is the honest measure of how often the agents were wrong.
What people decided
Proposed changes approved and rejected in the month, from the same approval rounds your team already works through.
How the month closed
The close readiness score and the business days from period end to close, shared with the Close Commitment so both pages agree.
What was caught
Anomaly findings raised and resolved, so a quiet month and an ignored month look different.
Where numbers come from
The share of journal entries that carry a source document, feed or run. Provenance you can click through to.
One link for the lender, the investor, the auditor.
Publish a read-only scorecard with a public link. It shows the grade, the cards and twelve months of history, never a transaction, a name or an amount. Revoke it any time.
Monthly snapshots
Each month is saved when it is first computed, so later edits to the books never rewrite the record.
Twelve months of history
The trend matters more than any one month. The history table shows whether the system is learning.
Ask Regent
The same figures are a question away in chat, and available to any MCP client with access to the business.
See your own books graded.
The free five-day pilot runs the readiness score on your real books by day four. The scorecard is part of the day-five close review.