Handle AI incidents calmly, from the first odd answer to the lesson learned
For risk, operations, compliance, customer and IT leads in organisations of 20 to 2,000 people that use AI: a 14-sheet workbook, 5 fillable forms, 7 brand-ready Word files, 4 cheat sheets and an e-book that follows one letting agency through its first eight AI incidents.
- A 14-sheet master workbook: AI register with kill-switch flags, 0-4 self-assessment and ranked gap plan, incident log that checks the 72-hour data clock, evidence register, decision log, business case and dashboard
- Severity levels, a first-hour card and a four-rung containment ladder, with a one-page kill-switch and permission check for every AI agent
- A dated map of the reporting clocks an AI incident can start in the EU, UK, US, Canada, Australia, New Zealand and India, with official sources
- Twelve honest messages, three tabletop exercises, a 45-minute team briefing and a 14-question LMS-ready knowledge check
The AI Incident Response Playbook · Bodhih Pro Kits · 339 by Bodhih Training, trainers to 2,000+ companies since 2008.
How do you respond to an AI incident?
Acknowledge the signal, set severity, take the smallest action that stops new harm (revoke an agent's access rather than asking it to stop), keep evidence, tell your data protection lead in time for any reporting clock, explain honestly to the people affected, then run a blameless review. The AI Incident Response Playbook gives you the files for each step.
Want the full free guide first? Read AI incident response: how to detect, contain and learn from AI failures.
You already feel the problem. Here is what it is costing you.
Nobody knows it counts
Staff correct a strange AI answer and move on. The first ten signals disappear, and the incident is found by a customer or a journalist.
Nobody can stop the agent
The agent works at weekends; the people with admin rights do not. Typing 'stop' into a chat box is not a control.
Clocks start before anyone notices
An AI summary sends personal data to the wrong people, and a 72-hour reporting clock may already be running while the incident waits in a helpdesk queue.
Blame kills reporting
If the first reporter is treated as the problem, there will not be a second report.
Picture the next few weeks with the kit open beside you.
Imagine the morning a tenant writes 'your chatbot told me my deposit comes back in five days'. Your complaints handler knows exactly what to do: one line to the incident route, the exact words, the time. Within the hour your incident lead has set severity, blocked deposit questions in the chatbot and asked the vendor to preserve the logs.
By the afternoon your data protection lead has confirmed no personal data was involved, the holding note has gone to the lettings team, and the letter to affected tenants begins with 'we gave you wrong information, and we are sorry'. Ten days later the blameless review finds five causes and no culprit.
That is what this kit builds, in twelve weeks, with the workbook and templates you will open tonight.
22 files that do the work for you
The e-book shows you how. The tools make it happen. Here is exactly what you download.
The AI Incident Response Playbook (e-book, PDF)
The guide: methods, worked examples, scripts and exercises
Start here: AI incident response map
The job in nine steps with the file for each, your first hour, your first week, the 12-week plan and the 'brand it in five minutes' box.
AI Incident Master Workbook
14-sheet workbook: Settings, AI system register with kill-switch flags, 0-4 self-assessment with RAG, ranked gap plan, incident log with 72-hour clock MET/LATE checks, evidence register, decision log, business case, dated crosswalk, RACI, What changed log, glossary and a dashboard with 3 charts and traffic lights.
AI incident report form (anyone can use it)
Two-page fillable form for any employee to report a possible AI incident or near-miss in five minutes, with the facts the incident lead needs.
Severity and containment decision record
Fillable record for the incident lead: severity S1-S4 with criteria, the containment rung chosen, who approved it and the evidence kept.
Notification clock tracker
Fillable one-page tracker of every reporting clock an AI incident may start (data, cyber, AI Act, contracts, insurers), with start time, due time and who decides.
Blameless review record
Two-page fillable review: timeline from evidence, scope including silent affected people, five whys to system causes, what helped, fixes with owners and dates.
Agent kill-switch and permission check
Fillable check for each AI agent or customer-facing AI system: what it can touch, how to stop it in minutes, who can do it, and the date it was last tested.
AI incident response policy and procedure starter
Brand-ready one-page policy and eight short procedures (report, triage, contain, evidence, notify, communicate, review, practise) with placeholders.
Leadership sign-off and decision memo
One-page memo to adopt the playbook (part 1) and a reusable decision memo for S1 incidents and external reporting (part 2).
AI incident message library
Twelve ready-to-edit messages: escalation call script, staff holding notes, customer explanations, vendor preservation request, regulator holding note, board update and close-out.
Worked example pack: Fenwick Lettings
The fictional Fenwick Lettings case end to end: register, self-assessment, the chatbot and agent incidents, decisions, messages, the blameless review and week-12 results, with numbers that match the workbook.
AI prompt pack for AI incident work
Fourteen prompts for drafting timelines, five whys, customer explanations and tabletop injects, each with a human-review rule and a no-personal-data warning.
45-minute team briefing session plan and answer key
Minute-by-minute plan for briefing any team on spotting and reporting AI incidents, with slides outline, two exercises and the answer key for the 14-question knowledge check.
Tabletop exercise pack: three scenarios
Three 60-minute tabletop exercises (chatbot wrong information, agent loop, data exposure through a transcription tool) with timed injects, roles, scoring rubric and a debrief sheet.
AI incident first-hour card
One-page card for the incident lead: the first 60 minutes minute by minute, the five facts, the containment ladder and the clocks to check.
Six ways AI fails: plain-English card
Six failure types in plain English, mapped to OWASP's 2025 list, each with a real case, an early signal and the first containment move.
Numbers that matter: AI incident KPI card
Eight KPIs with definitions, how to calculate them from the workbook, sensible targets you set yourself, and the sourced figures worth knowing.
AI incident rules by market (November 2026)
Dated one-page summary of the reporting rules an AI incident may trigger in the EU, UK, US, Canada, Australia, New Zealand and India, with official sources.
AI incident knowledge check (LMS-ready)
14 scenario questions with four options, the correct letter and an explanation, ready to import into Bodhih LMS or any LMS.
AI incident 12-week plan and yearly cycle (calendar)
16 calendar events: twelve weekly programme meetings and four quarterly reviews.
AI incident task board (CSV)
Twelve programme tasks with owner, week and status, ready to import into Trello, Planner, Asana or Jira.
Chapter by chapter
Each chapter opens with a real situation, gives you a method and ends with something to do today, using the tools in your kit.
- Format
- PDF e-book + XLSX + fillable PDF + DOCX + printable PDF + ICS + CSV
- Files
- 22 files plus a one-click ZIP
- E-book
- 63 pages
- Works with
- Excel, Google Sheets, LibreOffice, Word, Google Docs, Acrobat Reader, Outlook, Google Calendar, Bodhih LMS and other LMSs
- Markets covered
- EU, UK, US, Canada, Australia, New Zealand, India (dated November 2026)
- Spelling
- British English with a US glossary
What counts as an AI incident
A working definition, near-misses, and six failure types in plain English mapped to OWASP's 2025 list.
Know what you run
Finding AI hidden in tools you already pay for; owners, tiers and kill-switch status in one register.
Score where you stand
A 0-4 self-assessment of ten practices, a ranked gap plan and a business case built from your own estimates.
Make it easy to report
One route, one five-minute form, the signals to teach, and monitoring that does not wait for complaints.
Triage and the first hour
Severity S1-S4, the first-hour card and the five facts in every escalation.
Contain the harm and stop agents safely
The four-rung containment ladder, kill-switch and permission checks, and restarting safely.
Keep evidence and know your clocks
What to preserve and the reporting clocks an AI incident may start in seven markets.
Communicate honestly
Four parts of an honest explanation, who hears what and when, and telling the people who did not complain.
Review, learn and practise
Blameless reviews, lessons digests, tabletop exercises and the dashboard.
Vendors, contracts and the AI you buy
Six contract questions, vendor follow-up and agents you build from parts.
Not just reading. Tools you will actually open again.
Start here
- 00-Start-Here-AI-Incident-Map.pdf: the job in nine steps with the file for each, your first hour and first week
Workbook
- AI-Incident-Master-Workbook.xlsx: 14 sheets including Settings, AI register, Self-assessment, Gap plan, Incident log with clock checks, Evidence register, Decision log, Business case, Crosswalk, RACI, What changed, Glossary and Dashboard
Fillable PDFs
- AI incident report form
- Severity and containment decision record
- Notification clock tracker
- Blameless review record
- Agent kill-switch and permission check
Word templates
- Policy and procedure starter
- Leadership sign-off and decision memo
- Message library (12 messages)
- Worked example pack: Fenwick Lettings
- AI prompt pack (14 prompts)
- Team briefing plan and answer key
- Tabletop exercise pack (3 scenarios)
Cheat sheets
- First-hour card
- Six ways AI fails card
- Numbers that matter KPI card
- Rules by market (November 2026)
Plan and training
- Knowledge check CSV (14 questions)
- 12-week plan and yearly cycle (ICS)
- Task board (CSV)
- The e-book
When you have worked through it, you will be able to…
Every outcome is practised with a worked example and a tool, not just described.
- List every AI system you run, with an owner, a risk tier and a tested way to stop it
- Give every employee a five-minute route to report a possible AI incident
- Set severity and act in the first hour with five clear facts
- Contain harm with the smallest effective action and stop agents by revoking access
- Map the reporting clocks for your markets with your data protection lead
- Explain incidents honestly and run blameless reviews that lead to owned fixes
- Practise with three tabletop exercises and track progress on a dashboard
Know your starting point. Prove your progress.
People who measure before and after improve faster and can show it. AssessAll, our sister assessment platform, has AI-graded assessments that pair with this kit. Take one now, work through the kit, then re-take it.
Signing It Off: AI Governance Judgment
Take the assessment →AssessAll assessmentData Governance & Data Quality Stewardship Judgment
Take the assessment →AssessAll assessmentDecision-Making Under Uncertainty
Take the assessment →JobularyTurn what you learn into a measured development plan (Jobulary)
Make it a growth plan →A taste of Chapter 6
The vendor's first suggestion on Monday morning was to type "stop creating purchase orders" into the agent's instruction box. Lewis had already done something better at 7.40am: he removed the agent's permission to raise purchase orders.
The honest comparison
| Free templates and typical advice | This kit | |
|---|---|---|
| Definition | 'Tell IT if something breaks' | A three-part definition with near-misses and six failure types |
| What you run | IT's list of four tools | A register with owners, tiers and kill-switch flags by formula |
| Agents | Ask the vendor to fix it | A tested permission check and time-to-stop for every agent |
| Clocks | Find out during the incident | A dated map for seven markets and an incident log that checks the 72-hour clock |
| Messages | Written under pressure | Twelve approved templates that never say 'the AI did it' |
| Practice | None | Three tabletop exercises with a scoring rubric, a briefing plan and a knowledge check |
Made for you if…
- Risk, compliance and operations leads who have just been told 'you own AI incidents now'
- Customer experience and IT leads running chatbots, voice bots or AI agents
- Data protection officers who need AI incidents to reach them in time
- Organisations using AI in the UK, EU, US, Canada, Australia, New Zealand or India
Probably not for you if…
- Anyone wanting legal advice: reporting decisions belong to your data protection lead and adviser
- Providers of high-risk AI systems needing a full EU AI Act post-market monitoring system
- Teams wanting model red-teaming, MLOps or technical security testing
- Anyone wanting an audit, certification or a guarantee that incidents will not happen
Built by trainers, tested in real rooms.
Bodhih Training has designed and delivered corporate programmes since 2008. Bodhih Pro Kits package what works in those workshops into a guide plus the exact tools we use, so you can apply it the same day without a facilitator in the room.
From here to using it in three steps
- STEP 1
Tap Get the kit
Name, email and country. UPI, card or netbanking in India; international cards everywhere else. No account to create.
- STEP 2
Download instantly
Your download page opens at once, with every file and a one-click ZIP. A permanent link lands in your inbox too.
- STEP 3
Open the first tool
Start with the quick-start chapter, then the main spreadsheet or planner. Most buyers use something from the kit the same day.
Your AI incident playbook, ready before you need it
Open the Start-here map tonight and put your first three AI systems in the register. Tomorrow, publish the reporting route. The next odd answer will reach the right person in minutes, not days.
Personal licence for the buyer's own use, including use within your own organisation and with its staff. Please don't share or resell the files. Team, company and consultant licences: solutions@bodhih.com.
What buyers say
No ratings yet. Only people who have bought this kit can rate it - every buyer gets a one-click rating link a couple of days after their purchase.
The Bodhih family, when you are ready for more
The EU AI Act Deployer Compliance Kit
Get an organisation that uses AI ready for the EU AI Act as a deployer.
The AI Literacy Programme Kit
Design, run and evidence a workforce AI literacy programme.
The AI Agent Deployment Playbook
Put AI agents to work safely in a team or small company, tool-agnostic.
AssessAll
250+ AI-graded assessments to measure skills and earn verified credentials.
Jobulary
Individual development plans that start from evidence and stay yours.
Bodhih Training
Run this topic as a live workshop for your team, in person or online.
Pewple
Book a vetted coach for career, leadership or communication. Pay per session.
Bodhih.org pathways
Certification pathways in Applied AI, Workplace English and management.
The AI Incident Response Playbook: FAQs
Read the free guideAlready bought? Resend my linkWhat is an AI incident?
An event where an AI system you use caused, or nearly caused, harm to people, their rights, property or your organisation, or did something its owner would not approve. Near-misses count. Chapter 1 gives a three-part definition and six failure types.
Do I need an AI incident response plan if we only buy AI tools?
Yes. Most incidents in organisations that buy AI start with a vendor's system, but the harm lands on your customers and staff, and you remain responsible for what your tools tell people. The kit is written for organisations that use AI rather than build it.
Do I have to report AI incidents under the EU AI Act?
Providers of high-risk AI systems report serious incidents; deployers who identify one must inform the provider. These duties apply with the high-risk rules, from 2 December 2027 for Annex III systems after the Digital Omnibus on AI. Personal data breaches have separate clocks under GDPR. Your adviser decides what applies to you.
Is there an AI incident reporting law in the UK or US?
At November 2026 there is no general AI incident law in the UK, and the US federal AI Incident Reporting Act is a bill. Data protection, breach-notice and consumer laws still apply. The Rules by market sheet has the dated position and official sources.
How do I stop an AI agent that is misbehaving?
Revoke its permissions or use a tested kill-switch. Do not rely on telling it to stop. The Agent kill-switch and permission check gives you one page per agent and a time-to-stop to improve.
How long does it take to set up?
The kit's plan runs over twelve weeks at about 140 hours of the incident lead's time, plus a 45-minute briefing for staff. Your first hour gets the register started and the reporting route published.
Can I do this without a consultant?
Yes for the playbook itself: the kit is written for a non-specialist lead. Bring in your data protection lead and legal adviser for reporting decisions and contracts.
How is this different from a cyber incident response plan?
AI incidents often start outside IT (a complaints inbox, a recruitment shortlist) and include harms that are not breaches, such as wrong information or unfair outcomes. The playbook connects to your cyber and data breach processes for the parts they cover.
Does the kit include training for staff?
Yes: a 45-minute briefing plan, three tabletop exercises and a 14-question knowledge check as a CSV you can import into Bodhih LMS or any LMS.
Which frameworks does it use?
It uses the OWASP Top 10 for LLM Applications 2025 to name failure types and the OECD's common reporting framework for AI incidents to shape log fields, both paraphrased. Not affiliated with or endorsed by OWASP, NIST, ISO, IEC, the OECD, the European Commission or any regulator named.
How can my team measure their AI governance judgement?
AssessAll offers AI governance and data governance judgement assessments, and Jobulary turns the results into a development plan.
Will the kit be updated?
It was reviewed in November 2026. Dated content sits on two pages marked 'Correct at November 2026', and the workbook's What changed sheet helps you track changes each quarter.
How much does The AI Incident Response Playbook cost?
₹1,999 including 18% GST for buyers in India, or $49 for buyers outside India. One-time payment, no subscription, with an automatic invoice.
How do I get the files after paying?
Instantly. Your download page opens the moment payment succeeds, and a permanent download link is emailed to you. Download each file separately or everything in one ZIP.
What software do I need?
Any of the usual tools: Excel or Google Sheets for spreadsheets, Word or Google Docs for templates, Adobe Acrobat Reader, Preview, Xodo or GoodNotes for the fillable PDFs, and Google, Outlook or Apple Calendar for the calendar plan.
Can I get a refund?
Because it is a digital download, kits are not refundable once downloaded. We always refund a duplicate charge or a failed delivery, and if a file is corrupted and we cannot fix it within 3 working days. Full terms: bodhih.org/refunds.
Can my company buy this for a team?
Yes. Team and company licences, and workshops on this topic run by Bodhih Training, are available. Write to solutions@bodhih.com or call +91 99000 11601.






