Over 25 years of leading WASH responses across South Asia, the Horn of Africa, the Middle East and beyond, I’ve seen technology both accelerate life-saving work and create new risks when used carelessly. Artificial intelligence is now entering our sector at real speed. From coordinating large multi-partner WASH responses in Tigray, Amhara, Gaza and the West Bank, Cox’s Bazar, South Sudan, and during the El Niño drought and cholera outbreaks in Ethiopia, I’ve started to explore practical ways field and coordination colleagues can put AI to work — without losing the human judgement that remains at the heart of what we do.
But as we adopt these tools, one issue has become impossible to ignore: **data privacy**.
What I’m Already Seeing AI Deliver
In the responses I’ve led, AI is beginning to help with the tasks that have always consumed time and scarce analytical capacity:
- Rapid damage and needs assessment using satellite and drone imagery (tools such as SKAI and DEEP can flag damaged water infrastructure or flooded areas in hours rather than weeks).
- Predictive analytics for early warning — platforms like HungerMap Live combine climate, market, conflict and nutrition data to highlight emerging food and WASH crises before they peak.
- Language and information management support: summarising long situation reports, translating technical guidance, and pulling key points from partner updates.
- Operational efficiency: drafting donor briefings, flash appeal text, risk analyses and meeting notes.
- Community-facing tools such as chatbots that help deliver verified hygiene messages or referral information in multiple languages.
In WASH Cluster and sector coordination roles, the most immediate value has been processing the volume of partner reports, gap analyses and donor materials that arrive every week.
Practical Ways I (and Field Colleagues) Can Use AI Day-to-Day
Drawing from the reality of coordinating 40–100+ partners under pressure, here is how I see AI fitting into daily work:
- Situation analysis and reporting: When preparing the WASH component of a response plan or flash appeal (as I did in Northern Ethiopia and Palestine), secure generative AI tools can summarise partner reports, extract recurring gaps, and produce a first draft of needs and priorities. The output is never final — it is a starting point that the team then verifies against field data and partner knowledge.
- Donor engagement and resource mobilisation: In Ethiopia I mobilised over $200 million across two years through intensive donor briefings. AI can help draft concise talking points, tailor messaging to different donor priorities, and organise the evidence base more quickly. The relationship-building and strategic judgement remain firmly human.
- Capacity building and localisation: A recurring priority in my recent UNICEF roles has been strengthening local partners. AI can help generate simple training materials, checklists on WASH standards or anticipatory action, and translated guidance that local organisations can adapt. In low-connectivity settings, we still need offline or lightweight options.
- Risk-informed and climate-resilient programming: For flooding, cholera and conflict-related displacement responses, predictive models and scenario analysis can support the risk analysis we already do. They complement, rather than replace, community knowledge and partner assessments.
- Everyday coordination tasks: Meeting notes, action trackers, draft emails to government counterparts or municipalities, and initial analysis of qualitative feedback from hygiene promotion activities can all be accelerated. This frees capacity for the face-to-face coordination and advocacy that matter most.
Data Privacy: The Non-Negotiable Foundation
Every one of the uses above involves data — sometimes highly sensitive data about people who are already extremely vulnerable. In my experience coordinating responses in conflict and displacement settings, I have seen how personal information (names, locations, family details, health status, displacement history) can become a protection risk if it falls into the wrong hands.
Recent high-profile breaches, including the 2026 cyberattack on the World Food Programme that exposed data of hundreds of thousands of Palestinian households in Gaza, are a stark reminder. When people are forced to choose between hunger and handing over personal details, “consent” is rarely free or informed. Once data leaves our control — whether through commercial AI tools, third-party vendors, or poorly secured systems — we can no longer guarantee it will not be misused.
From years of work in fragile contexts, I apply a few hard rules:
- Never put beneficiary data, protection-sensitive information, or unvetted partner reports into commercial AI tools such as public ChatGPT or Claude.
- Prefer organisation-approved, secure platforms with clear data-protection safeguards.
- Practise strict data minimisation: collect and process only what is truly necessary.
- Keep a human in the loop for any decision that affects people’s access to services or safety.
- Conduct proper risk assessments before introducing new digital tools, especially those involving biometrics, cash systems, or predictive analytics.
- Remember that local knowledge and community voice cannot be automated. The best AI use still starts from what communities and local partners tell us on the ground.
The ICRC Handbook on Data Protection in Humanitarian Action remains my key reference. It reminds us that data protection is not an obstacle to effective response — it is part of protecting people’s life, integrity and dignity.
Looking Ahead
In my work advancing HDP Nexus approaches, climate-resilient services and localisation, AI is becoming another tool in the kit — useful when it helps us reach more people with better-quality, more accountable WASH services, and risky when it creates a false sense of precision or distances us from the communities we serve.
I am still learning. Like many of my colleagues, I experiment carefully, share what works, and keep asking the same question I have asked throughout my career: does this help us deliver more effectively, more accountably, and with greater respect for the people at the centre of the response?
That question now includes a second, equally important one: are we protecting their data with the same care we protect their physical safety?
That remains the only test that matters.