Call report analysis where every failed edit reads as one plain sentence, every swing ties to the ledger, and a person approves every explanation before filing
Every failed edit explained in plain words, tied to the ledger, approved before filing.
See one case, screen by screen ↓Call report analysis is the work of running the FFIEC edits on each draft Call Report, explaining quarter-over-quarter swings and writing the explanations for quality edits that still fail. Call Report Analysis, the solution, gives the regulatory reporting team seven agents that turn every failing edit into one plain sentence, split each swing into drivers tied to GL journals and subledger records, draft the explanation, and have a Skeptic agent try to break it. The analyst accepts, a manager approves, the Head of Regulatory Reporting releases and the Chief Accounting Officer signs.
AI use cases in banking: a buyer's guide to agentic AI for bank and credit union operations →
Updated
Call report analysis is the work between the edit run and the filing: every failed validity edit corrected, and every quality edit that still fails explained in writing before the agencies' Central Data Repository accepts the report. In the sample bank's FFIEC 031, each draft runs 1,906 edits, the first run fails 143, and after the fixes 38 still need a written explanation. They are written by a small team from spreadsheets, GL queries and emails to credit, treasury and deposits.
The FFIEC's Guidelines for Resolving Edits ask for the why, the amounts, the dates and the instruction reference. What gets written in the last week is often a sum that does not foot, a copied-forward explanation that no longer holds, or a phrase like “due to normal business activity”. The manager reviews a spreadsheet export, the review shrinks as the due date nears, and when the agency's analyst asks a question weeks later, the reasoning is rebuilt from email.
Estimated working days for a team writing about 38 explanations a quarter, by hand and with the agents: about 13 against about 4. Our arithmetic, kept modest.
Seven agents run the edits, split each swing, draft and challenge; analysts accept and managers approve.
Corvane National Bank, about $92 billion in assets, files the FFIEC 031 for September 30, 2026. It is Thursday, October 22, day 22 of 30, and the report is due Friday, October 30. Here is one failing edit, screen by screen, in the working solution.
Nadia opens the explanation docket: 38 explanations needed, 29 ready, 9 left, and seven of the nine are hers. Each failing edit is one sentence with its figures, such as “Nonaccrual CRE loans on Schedule RC-N rose 64% from June ($48.2M to $79.1M).” Under each one sits the agents' draft or the Skeptic's challenge, with record chips, the owner and a status chip: Drafted, Challenged or Ready.
Draft 3 ran 1,906 edits: 1,766 FFIEC validity and quality edits and 140 house checks, with 0 validity edits failing. Failures fell from 143 on Draft 1 to 51 on Draft 2 to 38. Of the 143, 12 validity edits were corrected, 61 were corrected in the data, 32 cleared after reloads, and 38 need a written explanation.
The Edit agent never changes a reported figure and never marks an edit resolved.
Nonaccrual loans secured by other nonfarm nonresidential properties moved $30.9M, from $48.2M at June 30 to $79.1M at September 30. The first draft names three suburban office loans placed on nonaccrual August 14, Halvorsen Park Office ($17.6M), Brightwater Commons ($11.2M) and Larchmont Corporate Center ($6.4M), and they add to $35.2M. The residual is −$4.3M, above the $0.25M limit, so the Skeptic challenges the draft and proposes the fix.
“Sum does not foot.” Peer trends sit beside the item, labeled “Context only, never a driver”.
Nadia clicks the citation. Core payment record PMT-2026-0922-118, posted September 22, 2026, applied $4,300,000 of principal to loan LN-41230 from the proceeds of an outparcel sale, leaving $2,100,000. The passage the citation points to is highlighted, and opening the record is logged.
Nadia asks for a redraft with the Skeptic's finding and watches each step. The Variance agent decomposes again: $17.6M + $11.2M + $6.4M − $4.3M paydown = $30.9M, residual $0.0M. The Drafter rewrites to the FFIEC guidelines, with the why, the amounts, the dates and the instruction reference, and the Skeptic tries to break version 2.
Version 2 foots to the change: 716 of 1,000 characters, with numbered citations to 7 records. The Skeptic passed it at 0.96 confidence. Nadia accepts it and sends it for approval; Edit, and Reject with a reason code, were one click away.
“$30.9M increase: two office loans placed on nonaccrual August 14 ($28.8M) and one ($2.1M) after paydown.”
Marcus sees two explanations to approve, E-07 accepted by Nadia and E-15 accepted by Owen Pratt, each marked Skeptic passed and Residual $0.0M. He approves E-07, and the docket moves to 30 ready, 8 left. Below wait two corrections to reported figures: COR-061, $22.0M of public funds proposed by Sofia Lind, and COR-060, $3.4M of overdrafts he proposed himself and cannot approve.
Nadia selects the recurring explanations and accepts them in bulk. The Recurrence agent re-tests each one against this quarter's records first: deferred taxes and fiduciary income hold, but former branch premises in other real estate owned fell from $4.2M to $2.6M after a branch sale on September 18. Two go to a manager; one goes back to the Drafter.
The Recurrence agent never carries an explanation forward without a re-check.
Filtered to E-07, the trail shows the day-20 escalation, the Drafter's first version with its prompt version, the Skeptic's challenge citing PMT-2026-0922-118, the redraft, the pass at 0.96, Nadia's acceptance and Marcus's approval, each with its time and its agent version or sign-in. One export assembles it for examiners and internal audit.
All 7 release checks are met: 0 of 1,766 FFIEC validity edits failing, 38 of 38 explanations approved, 0 Skeptic challenges open, 61 of 61 figure corrections approved, every residual within $0.25M, and Schedule RC tied to the general ledger. Elaine Castellano, Head of Regulatory Reporting, releases the set on October 26, and Robert Hanley, Chief Accounting Officer, signs on October 27.
The target: filed on day 27 of 30, 3 days to spare, with 0 boilerplate explanations.
The same quarter's report, seen by five of the people who carry it from the edit run to the exam: what their quarter looked like, and what it looks like now.
Seven agents run the edits, split each swing, draft and challenge; analysts accept and managers approve.
Runs the FFIEC validity and quality edits from the current edit list and the bank's 140 house checks on every draft, and writes each failure as one plain sentence with its figures.
Splits each quarter-over-quarter and year-over-year swing into drivers from GL journals and the loan, deposit and securities subledgers, and shows the residual.
Writes each explanation to the FFIEC Guidelines for Resolving Edits, with the why, the amounts, the dates and the instruction reference, within 1,000 characters.
Tries to break every draft: foots every figure to a record, checks that dates fall in the quarter, and rejects incomplete and boilerplate wording, with its reason and the record.
Pulls public UBPR peer group trends for an item and shows them beside it, labeled context only.
Finds explanations that repeat last quarter's condition and re-checks each against this quarter's records before bulk acceptance.
Tracks every item against the 30-day window, escalates items unresolved at day 20, and assembles the release and examiner exports.
Accept, approve, release, sign. The agents propose; a named person decides.
Ask about any edit, swing or explanation in the September 30 Call Report, or say what to do next. Each answer cites its records and opens the screen it talks about.
Why did nonaccrual CRE loans jump this quarter?
$30.9M of the $30.9M is explained. Two suburban office loans went on nonaccrual on August 14 after new appraisals, $28.8M, and Larchmont Corporate Center went on at $6.4M but was paid down $4.3M on September 22. Peer office CRE rose 11% last quarter; that is context, not a driver.
Which explanations still use boilerplate?
One: E-12, C&I charge-offs, says “due to normal business activity”, which is on the banned list; the FFIEC guidelines ask for amounts and dates. Two more are challenged for other reasons: E-07 does not foot, and E-19 has a driver with no record.
Are we on track for October 30?
Day 22 of 30. 29 of 38 explanations are approved; 9 are left, 3 of them escalated at day 20. At the current pace the set is ready by October 24 and released on October 26.
Prepare the OCC examiner export for this quarter
It assembles every explanation with its versions, the Skeptic's challenges, who accepted and approved each, the evidence records and the edit runs, and asks you to choose what goes in.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds the agents work to and the standard the Drafter and the Skeptic share are settings. Each change is versioned and approved by a second administrator.
| Setting | Default | Choose from |
|---|---|---|
| Quarter-over-quarter swing | 10% and $5M | A percent and a dollar amount; the Variance agent decomposes every movement at or above both. Year over year: 20% and $10M |
| Residual limit | $0.25M or 2% | An explanation cannot be accepted while the unexplained change exceeds this |
| Skeptic pass threshold | 0.85 | Below it, the draft goes back to the Drafter |
| Redraft rounds before a person | 3 | After this, the Skeptic hands the item to the analyst |
| Escalation day | Day 20 | Items unresolved on this day go to the manager, with a copy to the Head |
| Explanation length | 1,000 characters | Sized to the explanation field in the regulatory reporting system; required parts: why, amounts, dates, instruction reference |
| Banned phrases | 8 | Six from the FFIEC guidelines' list of unacceptable comments, such as “Verified” and “Same as last quarter”, and the bank's own, such as “Due to normal business activity” |
| Retention | 7 years | Drafts, challenges, approvals and evidence links |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures and rules (FDIC FIL-44-2013: the Call Report is due within 30 calendar days of the report date, with no extensions, and is timely only if it passes the edits or explains the quality edits it fails; FFIEC Guidelines for Resolving Edits; FDIC: 4,421 insured banks at the second quarter of 2025). Corvane National Bank and the people, properties, loans and records named on this page are fictional: characters and sample data in the working solution.
Call report analysis is the work of running the FFIEC edits on each draft Call Report, explaining quarter-over-quarter swings and writing the explanations for quality edits that still fail. Call Report Analysis, the solution, gives the regulatory reporting team seven agents that turn every failing edit into one plain sentence, split each swing into drivers tied to GL journals and subledger records, draft the explanation, and have a Skeptic agent try to break it. The analyst accepts, a manager approves, the Head of Regulatory Reporting releases and the Chief Accounting Officer signs.
The Call Report must reach the agencies' Central Data Repository within 30 calendar days of the report date, with 5 more days for an institution with more than one foreign office, and no extensions. It counts as on time only if it passes the validity and quality edits or contains explanations for the quality edits it does not pass. The solution keeps the due date in view and escalates items unresolved at day 20 to the manager.
The Drafter writes to the FFIEC Guidelines for Resolving Edits: why the condition is not an error, with the amounts, the dates and the Call Report Instructions reference, within 1,000 characters and every figure cited. The Skeptic then foots every figure to its record and rejects the wording the guidelines call incomplete, such as “Verified” or “Same as last quarter”, and the bank's own banned phrases. It always shows its reason and the record.
Validity edits are corrected, not explained; the FFIEC guidelines make no provision for written explanations of them. The Edit agent never changes a reported figure: a correction is proposed by a person and approved by a manager who did not propose it, and the release checks require 0 validity edits failing.
Yes, in bulk, but only after the Recurrence agent re-checks each one against this quarter's records. In the sample quarter two of three held, and one, former branch premises in other real estate owned, no longer matched after a branch sale, so it went back to the Drafter.
No. The agents draft, decompose, challenge and escalate; they never change a figure, accept, approve, release or submit. An analyst accepts each explanation, a different person approves it, the Head of Regulatory Reporting releases the set to the regulatory reporting system, and the Chief Accounting Officer signs. Every step is on the audit trail with who, what, when, why and the evidence.
Explanations go to the agencies, so they name properties and loan numbers, never borrowers. Loan-level records are visible to reporting roles only, the agents have read-only access to every system of record, BSA case data is excluded from evidence, and records are kept for 7 years.
The Agentic Solution Engine builds and deploys it from your requirements and documents, such as your house checks, thresholds, explanation rules and a few past quarters' explanations, and it goes live once every quality gate has passed. We will walk you through it on your own last Call Report first.
The 5300 ready for the CFO in three days, every figure traced to the ledger.
1-min demoFR 2052a Liquidity ReportingEvery FR 2052a table tied to the ledger and every reclass cited before the ALCO sees it.
1-min demoRegulatory Reporting ReconciliationBreaks found and fixed at the source before filing, not after an examiner asks.
1-min demoWe'll run Call Report Analysis on your last quarter's draft, edit results and explanations.