AI Incident Response — Part 2: Recovery, Post-Incident Review & Readiness — 42 Templates, Checklists & Worksheets for Eradication, Staged Recovery, the Review, the Register, Exercises, On-Call Design & the Metrics That Matter

Containment stops the bleeding; it heals nothing — and quiet is not a condition, it is an absence of observed noise. 42 production templates, checklists & worksheets distilled from Chapters 8–12 and Appendices A–C of AI Incident Response: the eradication walk and the verified vendor pin, staged re-enablement (shadow, canary, partial, full) under a regression watch, the evidence-led review with no names in the cause column, the corrective-action register that closes only on verified effect, tabletops and dirty injects, on-call pairs and the second phone, the four clocks and the recurrence rate — ending in the book's 44-line Part 2 gate, its 28-line Master Checklist, and its full Template Library. PDF + Markdown source.

The systems were quiet on the second day. The causes were still sitting where the night had left them.

Containment is the dramatic half of an incident, and it is the half most teams practise. Then comes the morning after: the wrong fact still resident in a cache nobody checked, the vendor model still unpinned, the review postponed until the evidence has rotated away, the agreed repairs dying quietly of no owner, no date, and "done" defined as effort spent — until the same failure arrives again, on schedule, and is debated all over again in the first hour. AI Incident Response — Part 2 is the machinery for that second half: 42 copy-paste templates, checklists, and worksheets distilled from Chapters 8–12 and Appendices A–C of the dedicated incident-response book — eradicate the cause everywhere it lives, return the system to service through stages each earned by evidence, review the incident while the evidence is still warm and the cause column carries no names, close every repair only when its effect has been measured, practise the whole response before it hurts, and staff the night so the next 03:12 is unheroic.

Every item carries the book's tested positions built in: an eradication walk ordered along the path an answer travels — sources, indexes, caches, prompts, vendor configuration, the sibling system — with the negative proved per location and the incident itself replayed as the decisive test; a vendor pin verified in configuration, because a pin described in a message is a claim and a pin visible in configuration is a fact; staged re-enablement through shadow, canary, partial, and full, each stage entered on a test and held for a window spanning the conditions that produced the incident, under a step-back rule any watcher can invoke instantly without a meeting, guards retired one at a time and the harm-adjacent guard last; closure on five evidenced criteria, never on quiet, because quiet is the baseline state of an unexamined system; a review held within days, facilitated independently in a small room of contributors, built on a timeline from timestamped records with its gaps labelled and its awareness points kept separate, findings written as condition, evidence, effect, and repair direction so a stranger could act on them; a corrective-action register that is the single home for every repair, with priority set from severity, verification methods written at entry, closure only on measured effect over a window by someone other than the owner, failed verifications reopened in the same entry at higher priority, and an aging report that escalates lateness in the open; exercises in three forms chosen by the question they answer, scenarios built mostly from your own history with at least three forks, injects that arrive dirty and on the clock, and scoring against the four clocks, the artifacts, and the fork decisions — never against confidence; on-call rotations in primary–secondary pairs with automatic promotion when a primary goes silent, a rostered second phone for communications and compliance with waking thresholds agreed in daylight, runbooks held to one bar — executable by a tired, competent stranger — and the four clocks defined by fixed endpoints, published after every incident, with recurrence classified by pattern as the honesty metric the programme cannot flatter. Where the book gives an example — its review rhythm, its exercise cadence, its stall triggers, its worked week of recovery — the template labels it as the book's example, illustrative for a mid-size operator, for you to refill with your own committed values. Bracketed fields are yours to fill; nothing here invents a figure for your organisation, and where the book itself states no figure — its corrective-action class limits are the case in point — the template gives you the fill-in, not a number dressed up as the book's. The book's companies and systems — Lakeshore Health Network, the support drafting assistant Scribe, the messaging assistant Clara, the model vendor Copperline AI — are fictional composites, labeled fictional in the book itself. One honesty note, in the open: the book's Appendix B quick reference reprints the severity matrix in compressed values that contradict the book's own Chapter 2 matrix. Template 40 carries the appendix exactly as printed and flags the conflict in the open, with the instruction to reconcile and post one matrix — it does not silently harmonise the book against itself. This pack is incident-response machinery, exactly as the book frames it: it is not legal advice, it gives no regulatory-filing guidance, and it guarantees no outcome.

What's inside — 42 templates, checklists & worksheets in 8 sections

1. Eradication & the Vendor Pin — Find the Cause Everywhere It Lives (6) — Eradication Walk Sheet — the whole path an answer travels, location by location · Incident Replay Test Sheet — the decisive test is the incident itself, replayed · Vendor Change Facts Sheet — establish the change in writing, early on the second day · Vendor Pin Verification Sheet — a pin in a message is a claim; a pin in configuration is a fact · Vendor-Side Eradication Sheet — fold the change into the causal account · Eradication Record — prove the negative per location, including the clean checks

2. Staged Recovery & the Regression Watch — Return Through Earned Steps, Close on Criteria (6) — Staged Re-enablement Table — shadow, canary, partial, full, carried whole · Shadow & Canary Entry-Test Sheets — prove it unseen, then prove it on one briefed slice · Partial & Full Stage Sheets — widen one notch at a time; retire the harm-adjacent guard last · Step-Back Rule Card — any stage goes back down instantly, without a meeting · Regression Watch Build Sheet — look for this harm, on purpose, staffed by named people · Closure Criteria & Clara's Week Back — close on criteria, never on quiet

3. The Review That Repairs — Post-Incident Review Without Blame Theatre (6) — Review Timing & Evidence-Preservation Sheet — hold it while the evidence is still warm · Room & Facilitator Sheet — a small room of contributors, an independent pen · Evidence Timeline Build Sheet — built from records first, with its gaps labelled · Contributing-Factors Tree Worksheet — factors are plural, systemic, and evidence-tied · No Names in the Cause Column Rule Sheet — owners in the actions, never people as causes · Review Elements Table, Findings Form & the Review Skeleton — written so a stranger could act

4. Actions That Close — the Corrective-Action Register (6) — Why Actions Die — Diagnosis Sheet: no owner, no date, effort as done, scattered homes · Corrective-Action Register — Fields Table, carried whole · Register Entry & Read-Back Discipline Sheet — one home, honest states, visible to the room · Priority Classes Sheet — P1, P2, P3, tied to severity, never to the room · Verification-Is-Closure Protocol — proposed is not done, and done is not verified · Aging Report, Reopen Rule, Register Metrics & the Four Worked Actions

5. Practice Before It Hurts — Tabletops, Game Days & Injects (6) — Three Forms Decision Sheet — choose the exercise by the question it must answer · Scenario Design Sheet — a controlled reproduction of a failure pattern, with forks · Scenario Library Table — the programme's memory, carried whole · Inject Design & Delivery Sheet — surprises that arrive dirty, on the clock · Scoring Sheet — the four clocks, the artifacts, and the fork decisions · Seats, Facilitation & Register-Conversion Sheet — the cast is the curriculum

6. Ready at 3 A.M. — On-Call Design, the Second Phone & Runbooks (5) — Primary & Secondary Rotation Design Sheet — staff the behaviour users experience · Second Phone Roster & Waking Thresholds Sheet — the institutional response, at incident speed · Runbook Quality Bar Sheet — executable by a tired, competent stranger · Stall Escalation Ladder Sheet — escalation climbs itself on the clock · One Year Later — Worked Readiness Example Sheet: fast, unheroic, and truthful on the same page

7. The Metrics That Matter — Four Clocks, Recurrence & the Readiness Review (5) — Four Clocks Definitions Sheet — fixed endpoints, published after every incident · Readiness Scorecard & Metrics Table — speed, honesty, and preparation, read together · Recurrence — the Honesty Metric Sheet: classified by pattern, never flattered · Readiness Review Agenda Sheet — the scorecard read as a whole, on a fixed cadence · Appendix B Quick Reference — Severity Matrix & Four Clocks, Carried Exactly, With Its Conflict Flagged in the Open

8. The Book's Own Checklists & Template Library — Master Checklist, Templates & the Part 2 Gate (2) — Master Checklist & Template Library — Appendices A and C, Carried in Full · Part 2 Gate — Chapters 8–12 Sign-Off: the book's five chapter checklists, consolidated

Who this is for

Incident commanders and reliability leads who own the second day as well as the first hour, platform leads who own eradication, pins, and re-enablement, review facilitators and the accountable owners who must make repairs actually close, exercise designers and on-call architects building a standing readiness programme, and the operations executives who read the scorecard quarterly. The source book assumes you operate at least one AI system that touches customers, staff, money, or records, and that incidents — declared or not — are already part of your year. If your last incident's repairs are a meeting note and a good intention, start at Template 19.

What this pack is not

Not Part 1, and not dependent on it. Declaration and severity scoring, the first hour, containment, evidence capture, and disclosure drafting are the book's Chapters 1–7. They live in the Part 1 companion pack; nothing here restates them as working machinery, and this pack is complete and useful without them. Two deliberate exceptions, both flagged where they sit: Appendix A is the book's whole-book Master Checklist, carried in full as printed (its first three groups are one-line checks whose machinery is Part 1's — Template 41 says so), and Template 40 points to Chapter 2's severity matrix because Appendix B's reprint contradicts it — a pointer, never imported Part 1 content.

Not legal advice, and not regulatory-filing guidance. The source book's own position, carried in full: notification duties, regulatory clocks, and disclosure obligations vary by jurisdiction, sector, and contract, and must be confirmed with qualified counsel for your situation before being relied upon. Appendix C's regulator initial notice is carried as the book prints it, under the book's own instruction to confirm duties and wording with counsel before the incident, not during it. This pack builds recovery, review, and readiness machinery; it does not tell you what you must file, with whom, or by when, and it guarantees no recovery, review, or readiness outcome.

Not the observability pack. The companion AI Observability pack detects and monitors AI systems in production — telemetry, tracing, evaluation, drift, SLOs — and holds that book's incident runbook and postmortem item. This pack is a different book and a different discipline: it begins after the response, at eradication, and runs through recovery, review, and standing readiness. Its review is the response book's evidence-led repair review, not a contributing-factors postmortem template, and the two packs share no template.

Not the cross-book security slice. The companion AI Security Templates & Checklists pack holds the MCP-and-enterprise-agents security artifacts, including its own agent incident response plan and production-readiness sign-off. This pack ships no security control, audit-trail standard, vendor risk assessment, or production-traffic sign-off; its register, exercises, and readiness gate serve incident recovery and learning, not security compliance.

Not the governance machinery. The companion AI Governance in Practice packs build the standing portfolio machinery — register, tiering, approval lanes, policy, cards — and its Part 2 sibling carries incident governance as a policy topic. This pack is the operational discipline itself: the walk, the stages, the review room, the register, the drill, the rota. Different books, different artifacts.

No platform or tooling. The register fields, watch dashboard shape, and runbook bar are simplified, structurally honest templates. No incident platform, paging tool, register software, or dashboard ships here.

Companion products

The AI Incident Response — Part 1: Declare, Contain, Preserve & Communicate pack is this pack's sibling, built from Chapters 1–7 of the same book: a companion for the first-response half — never a requirement for this pack to work. The AI Observability pack is the detection-and-monitoring companion — the machinery that notices. The AI Security Templates & Checklists pack is the cross-book security companion. AI Governance in Practice — Part 1 and its Part 2 sibling are the standing-governance companions. This pack stands alone.

Note: the $99 Architecture Templates Bundle currently contains Products #1–#3 only. This pack launches standalone; no bundle inclusion is promised for this product.

Built by the author of AI Incident Response: The Runbook for Model Failures, Leaks, and Attacks, from the incident-response discipline behind production enterprise AI systems.

About the author

Naveen Kumar Josyula is the author of 35 AI books on Amazon KDP and the CEO and founder of AISoltrics Technologies Inc. He is the author of the source book, AI Incident Response: The Runbook for Model Failures, Leaks, and Attacks, from which every template in this pack is distilled.

Support promise

If anything in this file is wrong or unclear, email support@aisoltrics.com before you judge it — we fix verified issues within one business day and send you the corrected file free. (You can also reach the founder's office at aadhya@aisoltrics.com.) Updates to this product are free to past buyers via your Payhip library. Sold by AISoltrics Technologies Inc.