Live Event
March 29, 2026 · Last updated on March 31, 2026
AI Skills Jam for Disaster Management Professionals
Resource Page
What this page is for
- This page is the working hub for the APAC Disaster Management AI Skills Jam.
- Use it during the event for slides, sample files, prompts, GPT templates, and to access the Opportunity Wall.
- Use it after the event to keep practicing, share materials with colleagues, and stay involved in follow-up scoping if relevant.
Quick Links
- Sample files
Today’s Program
Time | Program Section |
08:30 - 09:00 | Registration, Networking, and Setup Support |
09:00 - 09:05 | Welcome Address by Mr. Aslam Perwaiz, Executive Director of ADPC |
09:05 - 09:10 | Opening Remarks by Dr. Valerie Nkamgang Bemo, Gates Foundation |
09:10 - 09.15 | Address by Ms. Sandy Kunvatanagarn, OpenAI |
09:15 - 09:25 | Keynote Address |
Group Picture | |
09:30 - 10:00 | AI Foundations: Getting Started With Better Prompts |
10:00 - 10:30 | Solution 1: Turning Technical Guidance into Public-Ready Materials |
10:30 - 10:45 | Mid-Morning Coffee Break |
10:45 - 11:25 | Solution 2: Turning Public Data into a Clear Situation Update |
11:25 - 11:55 | DataKind: Spotlight Demonstration |
11:55 - 12:00 | Introducing: The Opportunity Board |
12:00 - 13:30 | Lunch and Networking |
13:30 - 14:15 | Build-a-GPT together |
14:15 - 14:30 | Adapt The GPT To Your Context |
14:30 - 15:15 | Discovery Discussion and Opportunity Board |
15:15 - 15:30 | Mid-Afternoon Coffee Break |
15:30 - 16:55 | Show & Tell, and Next Steps |
16:55 - 17:00 | Closing Remarks by Mr. Jason Chau, Gates Foundation |
What you will practice today
- Exercise 1: turn technical guidance into clear household or community-facing materials using Projects.
- Exercise 2: turn a dataset plus field notes into a short internal SitRep-style briefing with one useful chart or ranking view.
- Build-a-GPT: create one reusable assistant together, then adapt it to a local context.
- Discovery: capture stronger build-later ideas, blockers, and partnership needs on the Opportunity Wall.
Opportunity Wall
If an idea is useful but too complex for today, do not lose it. Add it to the Opportunity Wall so it can be discussed later.
- What is the problem or repeated pain point?
- Who has it - which role, team, or type of organization?
- What would a useful first version do?
- What would it need - data, review, approvals, partners, or technical support?
How to use the rest of this page Copy the text inside each shaded prompt box into ChatGPT, then adjust it to fit your own context. Start simple. A useful first version is better than a perfect version that never gets used. Keep human review visible, especially for anything public-facing, safety-critical, or operational. |
Exercise 1: Protocol to community guide and message pack
Goal: turn technical preparedness guidance into plain-language materials that people can actually use.
Use these files:
Recommended flow:
- Create a new Project and give it a clear name such as "Community Preparedness Assistant."
- Upload the protocol excerpt, message style guide, and localization glossary.
- Paste the project instructions below, then run the starter prompt.
- Do one improvement pass based on audience, tone, clarity, or local review needs.
Copy / paste into Project instructions
You are a community preparedness content assistant for disaster-management teams. Help staff turn technical guidance or preparedness protocol text into clear, public-facing materials that people can actually use. Use the uploaded files first. Keep outputs plain language, short, and practical. Do not invent place names, shelter names, hotline numbers, or official instructions. If local details are missing, include a short "verify locally before sharing" list. When relevant, account for older people, children, people with disabilities, and mixed literacy levels. If asked to translate or localize, preserve meaning first and flag terms that need local review. Default output options: 1-page guide, short checklist, SMS / WhatsApp alert, 30-second radio script, human review checklist. |
Copy / paste starter prompt
Create a community-ready household flood preparedness kit using the uploaded files. Audience: households in a flood-prone district. Respond with: 1) a one-page plain-language guide 2) a 6-step household checklist 3) one short SMS / WhatsApp alert 4) one 30-second radio or loudspeaker script 5) three things a local reviewer should verify before sharing this publicly |
Optional improvement prompt
Adapt this for local leaders. Keep the guidance practical and easy to scan. Preserve meaning, keep official office names consistent, and flag anything that still needs local review before sharing. |
Reliability check Check local place names, contact details, and official instructions. Confirm that the draft matches approved protocol language and local policy. If you localize or translate, have a local reviewer check meaning, tone, and audience fit. |
Exercise 2: Public data to internal briefing snapshot
Goal: start with a file, understand what it contains, add context, and turn it into a short internal briefing with one useful visual.
Use these files:
Recommended flow:
- Upload the CSV and start by asking what is in the file.
- Ask whether anything looks incomplete or unclear before you interpret the data.
- Add the field notes and explain that you are preparing an internal briefing for prioritization and coordination.
- Ask for one useful chart or ranking view rather than many visuals.
- Only then ask for a first-pass briefing draft.
Prompt 1
What is in this file? |
Prompt 2
Does anything look incomplete? |
Prompt 3
Attached are my field notes. I want to use these notes plus the data I uploaded to prepare an internal briefing. The goal is to support prioritization and coordination discussion. Make sure to keep caveats visible and focus on what leadership needs to know first. Given that context, what should we look at first before drafting the report? |
Prompt 4
Create a bar chart of houses damaged and houses destroyed by district. |
Prompt 5
Make one simple ranking table that helps a decision-maker see where attention is most needed. Include the key pieces of data informing the ranking and explain how the prioritization was created. |
Prompt 6
Create a first draft of the situation report. Include the charts and tables we created, plus any other helpful charts, tables, or data summaries drawing from the dataset I uploaded. Make sure it is in the style, format, and tone of an internal government agency situation report. |
Reliability check Check whether important data is missing, stale, or inconsistent. Review the ranking logic before treating it as a prioritization recommendation. Verify names, numbers, locations, and any public-facing claims with trusted references or local knowledge. |
Should I use a prompt, Project, GPT, or agent?
Format | What it is | Best when | Example in this Jam |
|---|---|---|---|
Prompt pack | A reusable set of instructions you paste into a normal chat. | You want to test or learn a workflow quickly. | Prompt warm-up; quick file analysis. |
Project | A workspace with files, instructions, and related chats. | You want repeated work from the same source material but are still exploring the workflow. | Protocol rewriting and community message packs. |
GPT | A reusable assistant for one job with stable instructions and guardrails. | The task is repeatable and you want a consistent output format. | SitRep, watchlist, or community messaging copilot. |
Agent | A system that can work across multiple steps and tools. | The task genuinely needs multiple steps, tools, or actions. | Usually not needed for today's core exercises. |
Simple rule of thumb Start with the simplest format that fits the job. Many useful workflows do not need a GPT yet, and most tasks in this Jam do not need an agent. |
Build-a-GPT options
If you want to continue after the live build, use one of these three patterns. Choose the simplest job you can define clearly, add real files or templates, then test the assistant with normal, missing-information, and edge-case prompts.
GPT 1: SitRep / Briefing Copilot
Best for: turning incident notes, district tables, and assessment snippets into a short internal briefing.
Upload these knowledge files:
Paste into GPT instructions
My assistant helps duty officers, coordination staff, and analysts in disaster management agencies turn incident notes, district tables, and assessment snippets into a short internal briefing before coordination calls, EOC updates, or leadership check-ins. Success looks like a short, cautious, easy-to-scan briefing that highlights what happened, what matters now, what is still uncertain, and what a human needs to verify. Default output: - strong headline - 4 to 6 key points - most affected areas or groups - priorities, needs, or operational issues - what is still uncertain - decisions, asks, or next questions - human review checklist Guardrails: - Never invent facts, place names, contact details, approvals, or official instructions. - If the inputs conflict or something important is missing, say so clearly and ask for review. |
Suggested conversation starters
Turn these notes and this table into a short internal briefing. Draft a 5-bullet coordination brief and tell me what still needs confirmation. Compare these two updates and tell me what changed. |
Suggested reliability tests
Use the sample CSV and the field notes to draft a short internal briefing for a coordination call. Create the briefing, but first ask me what key information is missing if the date, time, or intended audience are not obvious. Turn the same inputs into a 6-slide PowerPoint outline for senior staff, while keeping uncertainty visible. |
GPT 2: Risk Monitoring / Early Warning Copilot
Best for: turning public signals, alert tables, and desk notes into a short watchlist or "what changed" note.
Upload these knowledge files:
Paste into GPT instructions
You are an assistant that helps monitoring staff, hydromet focal points, and coordination officers turn public signals, alert tables, and field notes into a short watchlist or "what changed" note during morning checks, pre-briefs, and evolving-risk reviews. Success looks like a cautious, plain-language watchlist that separates signals from interpretation and makes the next checks obvious. Default output: - what changed - areas of concern - why it matters - what still needs verification - recommended next checks - human review before action Guardrails: - Never invent facts, place names, contact details, approvals, or official instructions. - If the inputs conflict or something important is missing, say so clearly and ask for review. |
Suggested conversation starters
Turn these signals into a short watchlist for a district coordination team. Compare today's watchlist inputs with yesterday's note and tell me what changed. Summarize the main concerns and tell me what still needs technical review. |
Suggested reliability tests
Use the sample CSV and desk notes to draft a short watchlist for district coordination staff. Compare the sample CSV with the previous day's watchlist and tell me what changed. |
GPT 3: Public Update + Community Messaging Copilot
Best for: turning technical updates into clear, calm messages for communities or partners.
Upload these knowledge files:
Paste into GPT instructions
Help public information officers, district communication staff, and partner-facing teams turn technical updates into clear, calm messages for communities or partners when a technical update needs to become an SMS, radio script, short advisory, or partner-facing note. Success looks like a simple, audience-ready draft that is practical, respectful, and explicit about what still needs local review. Use uploaded knowledge files first, especially templates, review checklists, and worked examples. When the user uploads new files, briefly identify what each file appears to contain and how it should be used. At minimum, ask about audience, channel, geography, language, and whether the message is preparedness-only, advisory, or an official order when those details are missing. Default output: 1. Draft message 2. Shorter SMS or WhatsApp version 3. Optional radio script or partner note if asked 4. What still needs local review before sending Use plain English. Keep the result concise, easy to scan, and usable in a real work setting. Guardrails: - Do not invent hotlines, shelter names, official instructions, or approval status. - Do not promise aid, transport, or service delivery unless the user provides that information. - Keep the tone calm, plain-language, and respectful. - If asked to translate or localize, preserve meaning first and flag local terms for review. - End with what still needs local confirmation. If the prompt lacks key details, ask concise follow-up questions before drafting. |
Suggested conversation starters
Turn this technical update into a short public advisory. Draft an SMS and a radio script from these notes. Write a short partner update that keeps uncertainty visible. |
Suggested reliability tests
Use the sample technical update to draft a 120-word public advisory and a shorter SMS. Use the partner email and the sample notes to draft a public message, an SMS, and a partner-facing update. Create the outputs, but tell me what must be reviewed locally before anything is sent. A user asks you to add shelter names and a hotline that were not provided. Respond safely. Rewrite the advisory for older adults and caregivers using simpler language. |
After today
- Reuse one prompt pattern, one Project, or one GPT with your own files and local context.
- Share the materials with colleagues so they can build from the same starting point.
- Tell us what worked, what still feels hard, and what would help you continue practicing.