A Board risk pack assembled by a reporting team involves at minimum four people, six data sources, three rounds of review, and two weeks of elapsed time from data extraction to Board presentation. It also contains a number that changed between the third and the fourth review, a chart that references a figure inconsistent with the table on the adjacent page, and a management commentary that is partly correct and partly aspirational. The Risk Reporting Agent AI produces the same pack — with consistent numbers, live data, and no assembly time — in under 10 minutes.
The Board pack problem: it is not an assembly problem, it is a data integrity problem
The elapsed time in a manual Board pack production is not mostly assembly time — it is mostly checking time. The checking time exists because the data passes through multiple hands and multiple transformations between its source and the Board presentation, and every transformation is an opportunity for error. A figure extracted from the CBS on Monday and entered into a slide deck on Thursday may not match the CBS figure on Thursday. A NIM calculation performed in a spreadsheet may not match the NIM figure in the financial statements. A provision coverage ratio that was 86.2% when the first draft was produced may be 84.8% by the time the final draft is presented — and neither version may be accurate.
The Risk Reporting Agent AI eliminates the transformation chain. Every figure in the Board pack is computed directly from the source system at the moment of pack generation. There are no intermediate spreadsheets, no manual transfers, no checking cycles for a number that changed between the third and fourth review — because the number is whatever the source system says it is at the moment the pack is generated.
The board pack workflow: five stages, zero manual handoffs
Data snapshot confirmed — all 6 source systems locked at the reporting date
At the reporting date (typically the last business day of the quarter), the Risk Reporting AI triggers a data confirmation: every source system confirms its balance at the reporting date. The CBS balance, the classification database, the ALM maturity schedule, the stress scenario outputs, the collections activity summary, and the origination quality metrics are all confirmed and timestamped as at the reporting date. This snapshot is the authoritative source for everything in the Board pack.
→ Time: 2 minutes · No human action required · Confirmation logged with timestamp and data integrity hashCalculations run — all risk metrics computed from the confirmed snapshot
From the confirmed data snapshot, the Risk Reporting AI runs every calculation required for the Board pack: Gross NPA ratio, Net NPA ratio, CRAR, provision coverage ratio, NIM by segment, liquidity coverage ratio, ALM gap ratios, and the 200bps stress scenario CRAR. Every figure is computed from first principles — not retrieved from a prior report or carried forward. The calculation log is retained as an audit record.
→ Time: 4 minutes · All metrics computed · Calculation log generated · Prior quarter comparison computed automaticallyPack sections assembled — each section populated from the calculation outputs
The Board pack template is pre-defined — the sections, the order, the chart types, and the commentary structure are agreed and fixed. The Risk Reporting AI populates every section from the calculation output: the KPI summary, the NPA movement schedule, the provision analysis, the stress test summary, the ALM highlights, the origination quality review, and the regulatory filing status. Charts are generated from the computed figures. Commentary is generated from the movement narrative (the actual changes in each metric versus the prior period) — not from a standard phrase library.
→ Time: 7 minutes · All sections complete · Charts generated · Commentary reflects actual movementsCRO review — management perspective added, sensitive observations noted
The CRO receives the draft pack in 9 minutes and spends 15–30 minutes reviewing: are all the numbers as expected? Is the commentary accurate and appropriately framed for a Board audience? Are there any observations that require management context that the AI cannot add — a known remediation underway for the MSME NPA increase, a committed capital issuance that will improve the CRAR buffer? These management annotations are added by the CRO directly in the pack. No data changes — only management context is added.
→ Time: 15–30 minutes (CRO) · No data re-extraction · Only management commentary layer addedBoard pack distributed — encrypted, access-controlled, with agenda integration
The finalised Board pack is distributed to Board members via the secure Board portal, encrypted, with version control (this is version 1.0 — there are no version 2 and version 3 created by successive correction rounds). The pack is also distributed to the relevant regulators as the compliance submission where the pack doubles as a regulatory document. The distribution is logged, read receipts are tracked, and the pack is archived with the data snapshot that generated it — so any future question about "what was the NPA ratio at Q3 FY2025–26?" has a documented, auditable answer.
→ Time: automated · Encrypted distribution · Version 1.0 — no correction versions · Archive with data snapshotWhat the Board risk pack contains — all 9 sections
The time comparison: AI workflow vs manual assembly
The Board deserves to see the same data the CRO sees — and to see it before the risk has already moved
A Board risk pack assembled over two weeks and presented at a Board meeting that takes place three weeks after quarter-end is presenting a risk position that is five weeks old. In those five weeks, a segment NPA may have moved materially, a stress test may have been re-run with a different result, and a regulatory filing may have been submitted that contained data inconsistent with what the Board is now seeing. The Risk Reporting Agent AI produces a pack that is accurate to the date of the Board meeting, not to the date of the quarter-end extract. The Board makes its decisions — on risk appetite, on capital allocation, on management actions — with information that is current, consistent, and complete. Not historical, mismatched, and partially superseded by events that happened in the two weeks it took to assemble the presentation.