Can NGOs paste beneficiary data into AI tools?

No. This library is designed for anonymized, non-sensitive, or synthetic examples. Teams should not paste personal data, protection case details, medical information, survivor information, names, phone numbers, addresses, or raw beneficiary records into public AI tools.

Can AI make humanitarian eligibility or safeguarding decisions?

No. These prompts support drafting, checking, summarizing, and planning. Final decisions must stay with accountable people and established organizational procedures.

What should teams review before using an AI output?

Review accuracy, source limitations, missing context, bias, data-protection risk, conflict sensitivity, safeguarding implications, donor rules, and whether a qualified person has approved the final version.

Humanitarian AI prompt library

Useful AI prompts for NGO work, with human judgment still in charge.

Use these prompts for drafting, checking, summarizing, planning, and improving humanitarian work. Keep sensitive data out, verify every output, and treat AI as a disciplined assistant rather than an authority.

Grants MEAL Kobo / field data Reports Risk Policies

Safe use rules

Four gates before an AI output becomes organizational work.

1. Data

Use anonymized, non-sensitive, or synthetic material. Do not paste raw beneficiary records or protection case details.

2. Context

Add the country, sector, donor, audience, constraints, and assumptions only when safe to share.

3. Verification

Check facts, calculations, citations, donor rules, local terminology, bias, and missing field reality.

4. Approval

Let a responsible person approve the final version before it goes to a donor, community, partner, or public channel.

Copy-ready prompts

Start from a useful draft, then adapt it to your context.

Grants

Concept note pressure test

Use this before sending a concept note for internal review.

Act as a critical humanitarian programme reviewer. Review the concept note summary below for clarity, donor fit, evidence gaps, logic, feasibility, risk, accountability to affected people, and value for money. Do not rewrite everything first. Give me: 1. the three strongest parts, 2. the five most serious weaknesses, 3. missing evidence or assumptions, 4. questions a donor may ask, 5. a practical improvement plan. Context: [country/sector/donor/audience] Concept note summary: [paste non-sensitive summary]

Human review gate: Confirm donor rules, budget logic, and all claims with the programme and finance leads.

Compliance

Grant file checklist

Turn grant requirements into an internal action checklist.

Create a grant compliance checklist from the requirements below. Separate the checklist into: before signing, during implementation, procurement/finance records, MEAL evidence, reporting deadlines, visibility/communication, safeguarding or protection requirements, and close-out. Flag anything that needs legal, finance, donor, or senior management confirmation. Donor or grant requirements: [paste non-confidential requirements] Project duration: [dates] Team structure: [brief]

Human review gate: This is not legal advice. Validate against the signed agreement and donor guidance.

MEAL

Indicator quality review

Improve indicators before they become painful to collect.

Review the indicators below for clarity, measurability, feasibility, disaggregation, data source, collection frequency, denominator, risk of gaming, and usefulness for decision-making. Create a table with: original indicator, issue, suggested improvement, data source, collection method, and caution. Programme objective: [objective] Indicators: [paste indicators]

Human review gate: Confirm indicators with programme staff, MEAL staff, and community feedback realities.

AAP

Anonymized feedback themes

Summarize feedback without exposing people or cases.

Analyze the anonymized community feedback notes below. Do not infer identities. Group the feedback into themes, urgency level, programme area, possible root cause, suggested action, owner, and follow-up question. Separate complaints that may require safeguarding/protection referral from general programme feedback, but do not decide the case. Anonymized feedback notes: [paste only anonymized notes]

Human review gate: Follow your complaint, safeguarding, and referral SOPs. Do not let AI decide case handling.

Kobo / XLSForm

Survey question review

Find weak, leading, unsafe, or hard-to-code questions.

Review this draft field survey for humanitarian data quality and safe collection. Identify leading questions, double-barrelled questions, unclear response options, translation risks, unnecessary sensitive questions, skip-logic needs, consent issues, enumerator instructions, and data-cleaning risks. Return a table with issue, why it matters, safer wording, and whether the question should be removed. Survey purpose: [purpose] Draft questions: [paste non-sensitive questions]

Human review gate: Test with enumerators and context experts before any real collection.

Assessment

Rapid assessment plan

Shape a short, realistic assessment design.

Help me design a rapid humanitarian needs assessment plan. Use the context below and produce: key questions, secondary data to review first, sampling approach, minimum ethical safeguards, field team roles, data collection tools, daily quality checks, limitations, and a 72-hour analysis plan. Context: [safe context] Sector(s): [sectors] Time available: [time] Access constraints: [constraints]

Human review gate: Confirm access, security, informed consent, and do-no-harm concerns locally.

Reports

Donor report structure

Turn rough notes into a structured report outline.

Turn the non-sensitive implementation notes below into a donor report outline. Use clear headings for progress against outputs, evidence, challenges, adaptations, risk, safeguarding/accountability, financial or procurement notes to verify, lessons learned, and next steps. Do not invent numbers or success claims. Mark every missing fact as [VERIFY]. Donor format or headings: [if any] Implementation notes: [paste non-sensitive notes]

Human review gate: Verify every number, date, quote, output, and financial statement before submission.

Risk

Risk register starter

Create a first draft for human review.

Create a practical risk register from the project description below. Include operational, access, security, safeguarding, fraud, procurement, data protection, reputational, partner, financial, and delivery risks. For each risk, suggest likelihood, impact, mitigation, early warning sign, owner, and review frequency. Mark uncertain items as assumptions. Project description: [non-sensitive description]

Human review gate: Risk ratings and mitigations must be approved by accountable managers and context leads.

Training

Workshop session plan

Plan a practical learning session, not a lecture dump.

Design a practical training session plan for the topic below. Include learning objectives, audience assumptions, timing, activities, case exercise, reflection questions, materials, facilitator notes, accessibility considerations, and a short knowledge check. Keep the tone adult, practical, and respectful. Topic: [topic] Audience: [audience] Length: [time] Delivery mode: [online/in-person]

Human review gate: Adapt examples to the audience, language, culture, and organizational policies.

Policy

Policy gap review

Use this to improve structure before formal review.

Review the policy excerpt below for structure, clarity, missing controls, accountability, implementation practicality, training needs, records required, review frequency, and risks. Do not give legal advice. Return: strengths, gaps, unclear wording, operational risks, suggested sections, and questions for senior management or legal review. Policy type: [finance/procurement/HR/safeguarding/data/etc.] Policy excerpt: [paste non-confidential excerpt]

Human review gate: Policies need leadership approval and, where relevant, legal/safeguarding/finance review.

Partners

Partner readiness questions

Prepare respectful due-diligence questions.

Draft a respectful partner-readiness conversation guide for a local organization. Cover governance, registration evidence, programme experience, safeguarding, finance controls, procurement, MEAL, data protection, community accountability, staff capacity, and support needs. Use non-accusatory language and include a section on how the partnership can strengthen the organization rather than only extract information. Partnership purpose: [purpose] Country/sector: [safe context]

Human review gate: Do not treat this as formal verification or donor clearance. It is a readiness conversation.

Communication

Ethical story review

Improve public writing without exploiting people.

Review this draft public update for ethical humanitarian communication. Check dignity, consent risk, stereotypes, savior language, privacy, conflict sensitivity, overclaiming, donor-friendly exaggeration, and whether affected people are represented with agency. Suggest a stronger version that is clear, respectful, and evidence-based. Draft text: [paste draft without personal identifiers]

Human review gate: Confirm consent, images, names, location sensitivity, and organizational communication rules.

Prompt builder

Build a careful first prompt.

Task type Work area Audience Non-sensitive context

Generated draft

Choose a task type and add safe context. Your generated prompt will appear here.

Take AI Literacy

Need help using AI safely?

Request a short AI workflow clinic.

If your team needs help turning these prompts into a grant, MEAL, reporting, Kobo, or policy workflow, send a short non-sensitive request. It will enter the operations inbox as an AI Clinic item for follow-up.

Reference points

Built around responsible data and useful content.

IASC data responsibility guidanceSafe, ethical, and effective management of data in humanitarian action. OCHA Centre for Humanitarian DataData responsibility and harm-reduction principles for humanitarian operations. OCHA AI briefing noteSector-level discussion of AI opportunities, limits, and governance needs.