All AI tools, by sector
Every AI tool and initiative in the register (325), grouped by sector. Select a name to see its full assessment: cost, connection needs, languages and whether it is still running.
Access to Information 21
- #vBezpetsi
A chatbot from the Emergency Telecommunications Cluster, led by WFP, that helps people affected by the war in Ukraine find and apply for help from aid organisations through Telegram and Viber; the same chatbot was used earlier in Libya, Iraq and Ecuador.
- ALMA
ALMA is a free IRC virtual assistant on WhatsApp that helps refugees and other newcomers in the United States with jobs, housing and services, around the clock.
- Ask Save the Children
A Save the Children prototype that gives teachers, community workers and child protection staff context-specific guidance drawn from the organisation's own curated research, rather than from a general chatbot.
- CARM feedback translation and severity grading (Claude)
Mercy Corps' Community Accountability Reporting Mechanism (CARM) staff use Claude to translate incoming community feedback, tag it by theme and propose a severity grade, with a person confirming each grade. The API is used for production work such as…
- Cboard
Cboard is a free, open-source app that lets children and adults with speech difficulties communicate with picture boards and text-to-speech, in 46 languages, on the web and app stores.
- Clara (Disaster Response)
An AI chat assistant on the American Red Cross website that helps visitors find disaster relief, shelter, donation, volunteering and training information in plain conversation.
- CLEAR Global
CLEAR Global builds chatbots that give crisis-affected people health, refugee and emergency information in their own language, over Messenger, Telegram, WhatsApp, SMS and voice; examples include Uji (DRC) and Shehu (Nigeria).
- Digital Care Assistant (Rafiki)
An AI assistant on phones and tablets that gives SOS Children's Villages caregivers training, parenting advice and peer support without travelling to sessions, with a route to human staff.
- Digital Community Hub
NRC's Digital Community Hubs are cloud contact centres built on Twilio that let displaced people reach NRC by WhatsApp, web chat and phone; more than ten were set up in about a year.
- Disaster Tech Lab
Disaster Tech Lab, which sets up emergency connectivity in disasters, is developing an AI information hub that people in disaster zones can reach over local Wi-Fi without internet access; designed in Tech To The Rescue's 2024 AI bootcamp.
- FirstPass
Mobile app that connects refugees and humanitarian workers with volunteer interpreters in real time. AI matches requests to translators, and a generative AI feature (FirstPass) drafts translations that human translators then check. It fine-tunes language…
- GANNET AI
An AI platform from Data Friendly Space that answers humanitarian questions from verified sources with citations, summarises situations and reads documents in many languages; it has a free data explorer and paid organisational tiers.
- Jugalbandi
A WhatsApp chatbot that helps rural Indians find out about government schemes, such as scholarships and pensions, including who is eligible and what documents they need. People can type or speak in their own language, with AI4Bharat models handling speech…
- Kinan
An AI-powered fact-checking tipline/chatbot launched after the 2023 Turkey-Syria earthquake to curb false news in Arabic. It searches for verified information in reply to user questions and forwards claims it cannot answer to human fact-checkers.
- Refugee Solidarity Network
Refugee Solidarity Network built a proof-of-concept chatbot on a large language model to give asylum seekers in Turkey legal information, and published what it learned about accuracy, bias and oversight.
- SignpostAI
SignpostAI is an IRC-led programme building AI tools, including the SignpostChat assistant, that help Signpost's information services answer far more questions from displaced and crisis-affected people with human oversight.
- SignpostChat
A chatbot service giving information to people in Syria through digital help centres that are staffed and supported by local organisations. Run by a consortium of IRC, Mercy Corps and Internews. It relies on internet access, and the need for human…
- Sunbird UCI
Sunbird UCI (Unified Communications Interface) is an open-source configurable platform to enable multi-channel, personalised chatbot conversations and nudges across various channels such as WhatsApp, Telegram, SMS, email, etc. for governance use cases.
- U-Report SMS classification system
A machine learning text classifier that automatically sorts unsolicited SMS messages sent by young people to U-Report and routes them to the relevant UNICEF teams. It was built because up to 10,000 messages a week could no longer be read by hand.
- UNHCR
Digital petition boxes in Jordan where refugees leave text or voice messages for UNHCR, with AI helping to process and route them so more get an answer.
- UNICEF
A UNICEF U-Report chatbot that gives young people fleeing Ukraine safety advice, entry rules and information on support in neighbouring countries.
Cash & Voucher Assistance 8
- AI-targeted hurricane cash relief (Helene and Milton)
After Hurricanes Helene and Milton, GiveDirectly used Google's SKAI machine-learning tool to compare pre- and post-storm aerial imagery and find the worst-hit areas, then a second Google tool to narrow these to likely low-income households. People in those…
- Cash-transfer anomaly detection
UNICEF Yemen's system checks cash-transfer payments for irregularities before they are paid out, including duplicate payments, unusual timing and patterns that differ from a person's payment history, with a custom model for matching Arabic names. Every…
- Enterprise Deduplication Service
WFP's deduplication service uses names and photos to spot people registered more than once for assistance; staff review every flagged record and make the decision. It works without asking people to remove veils or turbans and uses open-source models. A…
- Flood
GiveDirectly uses Google's AI flood forecasts to send cash to households before floods arrive; during Cyclone Freddy in 2023 it reached more than 4,100 households in Mozambique, with expansion to Bangladesh and Nigeria planned.
- High-resolution ML poverty maps for Nigeria's COVID-19 Rapid Response Register
Deep-learning models trained on satellite imagery and other geospatial data estimate poverty at ward level across Nigeria, validated against the official Nigeria Living Standards Survey. The maps were used to guide which wards were included in the Rapid…
- International Rescue Committee (IRC)
The IRC uses Google's AI flood forecasts to trigger cash transfers before floods; in Adamawa State, Nigeria, about 3,000 households received cash ahead of the September-October 2024 floods, and the approach is being repeated in northern Kenya.
- MobileAid
MobileAid predicts household poverty from consented, anonymised mobile phone call detail records (call/SMS patterns, top-up behaviour, contact diversity, movement between towers) to pick likely extreme-poor households for cash transfers paid by mobile…
- Novissi machine-learning targeting
Togo's Novissi emergency cash programme used a two-step model to decide who received payments in its rural expansion. A deep-learning model on satellite imagery identified the poorest cantons, then an algorithm trained on a phone survey linked to…
Cross-cutting capability 57
- afri-mt5-base
African language generation and translation.
- afriberta_large
11 African languages including Amharic, Somali, Hausa.
- afro-xlmr-large
17 African languages, encoder.
- am-roberta
Amharic encoder.
- banglat5
Bengali generation.
- bert-base-arabertv2
Arabic understanding.
- bert-base-arabic-camelbert-mix
Arabic across dialects.
- bert-base-multilingual-cased-finetuned-amharic
transformers, pytorch, bert, fill-mask, endpoints_compatible, region:us
- chip-n-scale-queue-arranger
Orchestration pipeline (SQS/Lambda-based) for running trained machine learning models over satellite imagery at scale, tile by tile.
- data360-mcp
Model Context Protocol (MCP) server that gives LLM agents and chatbots direct access to the World Bank's Data360 Platform. Search, validate, and retrieve development indicators—from GDP and poverty to gender equality and climate—with structured metadata…
- DOZORRO AI risk detection
A machine-learning system that scans tenders in Ukraine's ProZorro public procurement system and flags those at high risk of corruption, such as collusion or unjustified disqualification. Flagged tenders go to civil-society monitors to check and report to…
- fastai-serving
Docker image for serving fast.ai models behind an API that mimics TensorFlow Serving, for deploying trained image classifiers.
- fasttext-language-identification
Language identification, 217 languages.
- FuzzyAI
A Python library for entity deduplication and cross‑dataset matching using fuzzy similarity for candidate generation, with optional rule‑based and LLM‑based validation and consolidation to resolve ambiguous matches
- hls-foundation-os
Open-source fine-tuning examples and training harness for the Prithvi HLS geospatial foundation model.
- IndicTrans2
An open-source machine translation model covering all 22 scheduled languages of India. It lets developers and organisations add translation between English and Indian languages to their own services.
- InkubaLM
A small (0.4 billion parameter) open language model trained from scratch mostly on African-language text, designed to run on ordinary laptops and CPUs. It can generate text and be fine-tuned for translation, sentiment analysis, named entity recognition,…
- Khaya
A translation app and web app that translates words and sentences between English and Ghanaian and other West African languages, with voice-to-text input. It comes out of Ghana NLP, an open-source initiative that publishes corpora, models and evaluation…
- label-maker
Prepares training and test data for satellite imagery machine learning by pairing OpenStreetMap labels with corresponding image tiles.
- LaBSE
Sentence similarity across 109 languages.
- Lesan
A machine translation system for Ethiopian languages that translates between English, Amharic and Tigrinya. It was built by an Ethiopia-based team to make web content available to people in their own languages.
- lonboard
Fast, GPU-accelerated interactive geospatial data visualisation in Jupyter notebooks, built on GeoArrow/GeoParquet for rendering millions of features.
- m2m100_1.2B
100-language many-to-many translation, larger.
- m2m100_418M
100-language many-to-many translation.
- madlad400-3b-mt
400-language translation.
- magasin
Cloud native open-source end-to-end data / AI / ML platform
- MARBERT
Arabic dialects, social media register.
- mms-1b-all
Speech recognition across 1,000+ languages.
- N-ATLAS
An open-source multilingual AI model for Nigerian languages that can transcribe speech and generate text. It was released mainly for developers, who are expected to build services such as voice and chat tools on top of it.
- nllb-200-1.3B
200-language translation, base checkpoint.
- nllb-200-distilled-1.3B
200-language translation, better quality.
- nllb-200-distilled-600M
200-language translation, smallest usable size.
- opus-mt-ar-en
Arabic to English translation.
- opus-mt-bn-en
Bengali to English translation.
- opus-mt-en-ar
English to Arabic translation.
- opus-mt-en-mul
English to many languages, same coverage.
- opus-mt-en-sw
English to Swahili translation.
- opus-mt-en-uk
English to Ukrainian translation.
- opus-mt-mul-en
Many languages to English, includes Tigrinya, Amharic, Somali, Lingala, Kinyarwanda, Haitian Creole.
- opus-mt-uk-en
Ukrainian to English translation.
- opus-mt-ur-en
Urdu to English translation.
- paraphrase-multilingual-MiniLM-L12-v2
Multilingual similarity, for deduplicating and clustering reports.
- pcn
Proof-Carrying Numbers (PCN): Trust is earned only by proof — the absence of a verification mark communicates uncertainty.
- Prithvi-EO-2.0
Geospatial foundation model for Earth observation, pre-trained on Harmonized Landsat Sentinel-2 (HLS) imagery; fine-tunable for downstream tasks such as flood mapping, burn scar mapping, crop segmentation and land-cover classification.
- raster-vision-aws
CloudFormation template for deploying Raster Vision (open-source computer-vision-for-earth-observation framework) batch jobs to AWS.
- REaLTabFormer
A suite of auto-regressive and Seq2Seq (sequence-to-sequence) transformer models for tabular and relational synthetic data generation.
- seamless-m4t-v2-large
Speech and text translation combined.
- Self-Supervised-Learner (Curator)
Trains an image classifier with far fewer labelled examples than normal by using self-supervised pre-training on unlabelled imagery first.
- Sunflower
An open-source conversational large language model built for Uganda's languages, which Sunbird says covers 67 African languages. It sits within a wider Sunbird stack (API, speech-to-text, text-to-speech, translation) that is already used in Uganda's Parish…
- tm-oam-ds-builder
Geo AI training dataset builder using tasking manager + openaerialmap & push to huggingface
- twitter-xlm-roberta-base-sentiment
Multilingual sentiment.
- whisper-large-v3
Speech recognition across 99 languages.
- whisper-medium
Speech recognition, smaller and cheaper to run.
- whisper-small
Speech recognition, runs on modest hardware.
- xlm-roberta-base-language-detection
Identify which language a message is in.
- xlm-roberta-base-ner-hrl
Named entity recognition, multilingual.
- zenml
ZenML 🙏: One AI Platform from Pipelines to Agents. https://zenml.io.
Early Warning & Disaster Preparedness 41
- ADA
Deep-learning tool from 510, the Netherlands Red Cross data team, that spots damaged buildings in satellite images after a disaster so response teams can plan where to go first.
- ADAM
WFP's ADAM automatically produces impact maps and reports within hours of earthquakes, tropical storms and floods, estimating people and infrastructure exposed; recent flood reports (Sudan, Chad) are public.
- CAST
ACLED's Conflict Alert System forecasts how many political violence events each country is likely to see over rolling four-week periods, updated monthly, as an early warning for organisations working in or on conflict.
- ChaBuD (Change detection for Burned area Delineation)
Deep-learning burned-area / wildfire change-detection codebase from the ChaBuD ECML-PKDD 2023 challenge, using paired Sentinel-2 imagery.
- Danish Refugee Council (DRC)
A Danish Refugee Council machine-learning model that forecasts internal displacement from past conflict, health, climate and food-security data, to help agencies plan ahead in the Sahel and East Africa.
- Danish Refugee Council (DRC)
A Danish Refugee Council model that traces how drought leads to displacement among pastoralists, from rainfall to pasture, livestock, prices and income, and lets users test responses such as emergency feed or cash.
- DisasterAWARE
Pacific Disaster Center's DisasterAWARE platform gives hazard warnings, impact estimates and situational awareness from thousands of data layers, for disaster managers.
- DisasterBot
An AI-assisted chatbot on WhatsApp, Telegram, X and Facebook that lets residents in disaster-hit areas send real-time reports, which are mapped on the free PetaBencana.id platform. Government agencies, responders and aid groups use the map to target response.
- Foresight
DRC's Foresight is a machine-learning model that forecasts forced displacement one to three years ahead for 26 countries from more than 120 open indicators on conflict, governance, economy, environment and population; forecasts are published in an annual…
- GEWFD
UNHCR's predictive analytics for early warning forecast border crossings and displacement so operations can prepare, for example to anticipate refugee arrivals in Brazil.
- Google Flood Hub
Google's flood forecasting service, which gives riverine flood forecasts up to seven days ahead. Free to use.
- HYDRAFloods
HYDRAFloods is an open-source Python tool from SERVIR-Mekong that produces surface-water and flood maps from satellite imagery using Google Earth Engine; documentation uses Cambodia as the example.
- IDETECT
IDMC's IDETECT reads news, UN and NGO reports and pulls out incidents that are likely to cause internal displacement, with their cause, place and numbers, in close to real time.
- IDP Migration Forecasting (Huynh & Basu)
A research model for forecasting internal displacement (IDP migration) in Syria, dated June 2020. It exists as an academic publication rather than a tool, and the source describes it as difficult for humanitarian actors to put into use.
- india-forecast-app
Runs wind and PV forecasts for India and saves to database
- International Committee of the Red Cross (ICRC)
An ICRC and Waseda University project that flies drones with thermal cameras and uses machine learning to pick out buried landmines, aiming for a cheap device humanitarian deminers can use.
- International Committee of the Red Cross (ICRC)
From October to December 2023 Google gave the ICRC daily flood forecasts and inundation risk maps for 20 priority locations in Somalia, used to plan operations and warn communities.
- International Federation of Red Cross and Red Crescent Societies (IFRC)
The IFRC Alert Hub is a free platform that gathers official hazard alerts in the Common Alerting Protocol format from authorities worldwide, with preparedness guidance for Red Cross and Red Crescent societies, media and emergency managers.
- Kuzi
An early-warning tool that uses machine learning on satellite, soil-sensor and weather data to predict desert locust breeding and migration and to map high-risk areas. Farmers and pastoralists get free SMS alerts two to three months before swarms are…
- MEACAM
An online platform that forecasts climate hazards, including floods, from satellite imagery and climate analytics, with forecasts updating several times a day. Run by Mercy Corps.
- MET Norway Weather API
Free digital access to high-quality, reliable and user-friendly 10-day weather forecast data for any geolocation in the world. Provides an open interface to a world weather data produced by MET Norway.
- metnet
PyTorch Implementation of Google Research's MetNet and MetNet-2
- Noor
Viamo's Noor is a voice assistant on ordinary phone calls, with no internet or smartphone needed, that answers disaster preparedness questions in colloquial Urdu in Pakistan.
- OCHA Typhoon Impact Prediction Model
A machine-learning model that predicts how much damage a tropical cyclone will do before landfall, from weather and geographic data, and triggers OCHA's anticipatory action funding; started in the Philippines and being extended to Fiji.
- PRISM
WFP's PRISM combines current weather and hazard data with historical baselines and data on vulnerable people, so governments can see climate risks building up and estimate impacts quickly; it is increasingly linked to social protection.
- Prithvi-WxC
Foundation model for weather and climate, pre-trained on MERRA-2 reanalysis data, with downstream tasks including short-range forecasting and climate model parameterisation.
- Productive Forests
Satellite Imagery Processing for Machine Learning Based Image Classification. Produtive Forests creates a forest inventory to support the sustainable use of non-timber related forest products
- Project Jetson
A UNHCR experiment that used machine-learning time-series models to forecast forced displacement in Somalia, so operations could plan ahead.
- SAADAAL Early Warning System
Shaqodoon's SAADAAL system uses AI to forecast droughts and floods and send alerts to farmers and partner agencies in Somalia; developed in Tech To The Rescue's 2024 disaster-management cohort.
- Sentry Syria
An early-warning system built on a network of acoustic sensors and human observers. It sent warnings to civilians through social media, local radio and sirens. Operated in north-west Syria from 2017 to 2024.
- skillful_nowcasting
Implementation of DeepMind's Deep Generative Model of Radar (DGMR) https://arxiv.org/abs/2104.00954
- Spotlight (Horizon Scan)
A UNICEF forecasting model that uses the GDELT news-event dataset to predict country-level changes in conflict events and deaths, feeding UNICEF's monthly scan of where humanitarian needs may rise.
- Start Network Automated Drought Model
Start Network is automating its drought risk model and adding impact and vulnerability data, so its early-warning and anticipatory funding can trigger faster; developed in Tech To The Rescue's 2024 cohort.
- Sunny Lives
SEEDS and Microsoft use AI on satellite images of roofs to score each house's risk from cyclones, floods, heatwaves and earthquakes, then turn the scores into advisories for the most exposed households in Indian cities.
- SUNO
SUNO, from ConnectHear and Ufone, turns disaster warnings into AI-generated sign-language videos and sends them free over a WhatsApp bot to deaf and hard-of-hearing people in Pakistan; funded by the GSMA Mobile for Humanitarian Innovation Fund.
- UNDRR
Early Warnings for All is the UN initiative launched in 2022 to put every person under an early warning system by 2027, organised in four pillars led by different agencies.
- UNHCR
Under Project Jetson, UNHCR explored computer vision on Landsat satellite imagery of Somalia to detect drought and relate vegetation conditions to displacement.
- University of Oxford
An open-source statistical package from the University of Oxford, with IDMC, that predicts where people will be displaced after earthquakes and cyclones, tested on more than 100 earthquakes in 38 countries.
- UNOPS
UNOPS built AI support for the whole mine-action cycle: RELand ranks areas for clearance from geospatial data, and satellite analysis tracks how housing, farming and jobs recover after clearance; used in Afghanistan and Colombia.
- veriflow
A verification pipeline for evaluating models and forecasts.
- WorldFloods
An Oxford-led AI model that maps floods in satellite images, small enough to run on board a satellite, built with ESA, Frontier Development Lab and partners including UNICEF; it was launched into space in 2021.
Education 1
- aprendIA
An IRC WhatsApp assistant that gives teachers in crisis-affected, low-bandwidth settings classroom guidance and professional development, combining training content with AI.
Education in Emergencies 5
- CADENA (Coordinación de Ayuda en Desastres en las Américas)
CADENA, a Mexican disaster-response organisation, is developing adaptive training modules that tailor disaster-response lessons to each volunteer; designed during Tech To The Rescue's 2024 AI bootcamp.
- Giga initiative
For UNICEF's Giga initiative, Development Seed trained models to find likely schools in satellite images, which human mappers then checked, to map unmapped schools so they can be connected to the internet.
- Hakeem
Hakeem is a chatbot from NRC and Microsoft that acts as a learning adviser, using machine learning to point young refugees to online courses that match their interests; it was tested with young people in Lebanon.
- Integrate
AI tool that gives practitioners guidance on integrating child protection with education in emergencies. It provides guidelines, tips, planning templates and checklists, and can help draft SOPs, concept notes and step-by-step guides.
- Rori
Rori is Rising Academies' AI maths tutor on WhatsApp that adapts practice and feedback to each pupil, used in partner schools in several African countries.
Environmental Monitoring 14
- airfortomorrow
UNICEF Open Source AI models to estimate PM2.5 Air Pollution in Thailand & Laos
- AirQo AI Platform
A network of low-cost air quality sensors with machine learning that calibrates sensor data and forecasts air pollution (PM2.5) for African cities. Data are shared through a free app, an API and open dashboards so cities, schools and families can plan…
- Defensa+AI
A tool built by SidLabs with Peru's Sociedad Peruana de Derecho Ambiental that sends real-time alerts about environmental crime in the Amazon to authorities and communities; a 2025 AI for Good award finalist.
- Dymaxion Labs Toolkit
The Dymaxion Labs Toolkit is a collection of open source libraries that leverage AI and Computer Vision to understand the physical world, by analyzing large scale geospatial data.
- e84-geoai-common
Contains Geospatial AI/ML related code
- EarthReach
EarthReach is a challenge from the 2025 edition dedicated to enhancing the accessibility of meteorological data visualisations produced by Earthkit, by equipping the plots module with LLM-powered alternative text generation capabilities.
- eo-learn
Earth observation processing framework for machine learning in Python
- GhostNetZero.ai
GhostNetZero.ai, from WWF, uses AI on sonar survey data to pinpoint where lost 'ghost' fishing nets probably lie on the seabed, so they can be recovered.
- graph_weather
Graph-based weather forecasting models. Originally, PyTorch implementation of Ryan Keisler's 2022 "Forecasting Global Weather with Graph Neural Networks" paper (https://arxiv.org/abs/2202.07575)
- Guardian
Rainforest Connection places solar-powered listening devices in forests; machine learning picks out sounds of chainsaws, poaching and species, and alerts rangers by phone. Recordings feed the Arbimon biodiversity platform.
- kmnlp-ai-backend
AI backend using LLMs and agents for the NOAA Knowledge Mesh NLP Prototype Project
- pybayts (BAYTS deforestation detection)
Python implementation of the BAYTS Bayesian time-series algorithm for near-real-time deforestation detection from satellite image time series (Reiche et al. 2015, 2018).
- Wildlife Insights
Wildlife Insights is an online platform where conservation teams upload camera-trap photos and machine learning identifies the animals, with maps and trend analysis.
- World Wildlife Fund (WWF)
WWF is piloting machine-learning algorithms that flag likely wildlife products in X-ray scans and shipping data at airports and seaports, with partners including TRAFFIC and Sandia National Laboratories.
Food Security & Livelihoods 24
- AgPile
A federated system, in development at CGIAR, for sharing agri-food data across organisations without pooling it in one place, so it can be used to build AI tools.
- AgriLLM
An open-source agricultural language model from the CGIAR AI Hub and AI71 that gives farmers, advisers and policymakers multilingual farming guidance.
- AI Agro
Open Source Precision Agriculture with Drones leverages a set of machine learning algorithms & computer vision developed with the goal of determining health states and relevant information for efficient management of resources of different types of crops…
- AI Genebank platform
A CGIAR platform that screens very large numbers of stored plant samples for climate-resilient traits and lets breeders query the results through a chat interface.
- ASIS
FAO's Agricultural Stress Index System uses satellite data, updated every ten days, to spot agricultural drought worldwide, with a country-level version and an archive back to 1984; it is registered as a Digital Public Good.
- ATIO
FAO's Agrifood Systems Technologies and Innovations Outlook (ATIO) is a knowledge base, co-designed with users, that catalogues technologies and innovations for food-system transformation.
- Catholic Relief Services (CRS)
CRS dashboards that pull global commodity prices and local market data into one place and use machine learning to forecast prices six to twelve months ahead, for country teams planning cash and market programmes.
- CGIAR
The CGIAR AI Hub, a shared digital space where CGIAR centres and partners pool data and build AI tools for agriculture so they work together rather than in silos.
- crop_productivity_package
Modular pipeline for monitoring agricultural productivity using MODIS EVI and Google Earth Engine, with support for cropland masking, zonal statistics, anomaly detection, and automated export.
- Darli AI
An AI farming adviser that farmers reach through WhatsApp or a conversational phone line built for low-resource languages. It teaches regenerative farming methods, tracks whether farmers apply them and measures results such as yields and carbon capture.
- FamPred
A research model that forecasts household food consumption 60 days ahead in Mali, Nigeria, Syria and Yemen from WFP's public hunger-monitoring data, using reservoir computing.
- FarmerChat
Digital Green's AI assistant that helps extension workers and smallholder farmers get crop- and location-specific advice in local languages by voice, text or photo.
- Food Systems Analytical Solution
Mercy Corps' Food Systems Analytical Solution is a cloud platform using machine learning and geospatial analysis to understand food systems; developed further in Tech To The Rescue's 2024 disaster-management cohort.
- HungerMap LIVE
WFP's HungerMap LIVE is a public website that combines food security, weather, conflict, economic and other data, with machine-learning estimates where data are missing, to track and forecast hunger in near real time.
- Kisan e-Mitra
A voice and text chatbot that answers farmers' questions about the PM-KISAN income support scheme, such as eligibility, registration and payment status. Answers are tailored to the farmer's own records and given in regional languages.
- Mercy Corps CROP-ID
Uses satellite imagery, machine learning and field validation to identify crop types. Named as a Syria case study in the source under 'recommended' rather than as deployed, so its status in Syria is unclear.
- MIRA
A CRS resilience-monitoring system in flood-prone districts of Malawi that combined household surveys by community enumerators with machine learning to predict which households would face food insecurity one to two months ahead.
- PAK-Poverty-Projections
In 2019, 21.9 percent of Pakistan’s population lived below the national poverty line. Since then, COVID-19, floods, and a macro-fiscal crisis with record inflation have severely affected incomes and economic activity. In the absence of new survey data,…
- Plantix
Plantix is a free phone app that diagnoses crop diseases and pests from a photo and suggests treatments, with a disease library and a farmer community.
- SmartCatch
WorldFish's SmartCatch lets small-scale fishers photograph their catch and have AI identify species, size and weight on a basic phone without internet; it won the 2025 AI for Planet prize.
- Statistics from Space
IFPRI, Seoul National University, the University of Twente and Eduardo Mondlane University built a machine-learning framework that uses satellite data to produce faster maize, soybean and rice production estimates for Mozambique's food-security planning.
- UlangiziAI
A WhatsApp farming adviser from Opportunity International that answers farmers' questions by text or voice in local languages, drawing on Malawi Ministry of Agriculture guidance through speech-to-text, translation and a language model.
- VIPS
Open platform for automatic pest prediction in agriculture created by NIBIO, Norway
- World Food Programme (WFP)
A WFP phone-survey tool that uses voice-to-text AI to record people's answers by voice, used to monitor what households eat in places that are hard to reach.
Health & Nutrition 29
- Africa Infodemic Response Alliance (AIRA)
A WHO-convened alliance of agencies and fact-checkers that tracks and counters health misinformation across Africa, using data analytics and machine-learning tools from partners such as UN Global Pulse; public output runs under the Viral Facts Africa brand.
- AI for simplified malnutrition treatment
The IRC is testing AI to support its simplified treatment for acute malnutrition, which it says is about 20% cheaper than standard treatment: photographing handwritten clinic records so they are digitised automatically instead of typed in by hand (with…
- AIRR
A Malaria Consortium pilot that tested whether AI can count a child's breathing rate from a phone video, to help health workers spot pneumonia in children under five.
- Amref Health Africa
Amref Health Africa, with Microsoft's AI for Good programme, built a model that forecasts malnutrition risk at sub-county level in Kenya from health records and satellite data, so partners and the Ministry of Health can act before cases rise.
- Antibiogo
Antibiogo is a free, open-source Android app from the MSF Foundation that reads photos of antibiotic susceptibility tests, so lab technicians without microbiology specialists can choose the right antibiotic; it works offline and is CE-marked.
- Awaaz-e-Sehat
A voice-driven mobile assistant that lets maternal health workers speak patient information and have it turned into an electronic medical record. Speech and text recognition modules fill in an Urdu record template from conversations held in different…
- BRAC-mPower pregnancy risk categorisation
A machine-learning model from BRAC and mPower that sorts pregnant women by risk of complications, so community health workers in rural Bangladesh know whom to visit first; it runs in an Android app that also works offline.
- CGM
Welthungerhilfe's Child Growth Monitor is a smartphone app that films a child and uses AI to estimate height and growth scores against WHO standards, to detect stunting and underweight quickly.
- CIMA
A smartphone app for parents in Zaatari camp, Jordan. It stores each child's vaccination record in Arabic or English and sends scheduled reminders before appointments. A 2019 non-randomised trial (471 children with the app, 465 without) found 24.6% of the…
- Clara (Blood Donation)
Clara, the American Red Cross chat assistant, also answers blood-donation questions on the Red Cross Blood website.
- EARS (Early AI-supported Response with Social listening)
A social listening dashboard that uses AI to spot trending and rising topics in public online conversations across many countries. NLP is used to infer whether posts are questions or complaints, so health authorities can detect rumours and information gaps…
- EIOS 2.0
WHO's Epidemic Intelligence from Open Sources scans public information, including automatically transcribed and translated radio, and from version 2.0 uses AI to flag signals of disease outbreaks for public health teams.
- Everwell Hub
The Everwell Hub is a best in-class patient management platform designed for high-burden public healthcare programs globally. The Everwell Hub supports the entire digital cascade of care from diagnosis through treatment support and completion, including…
- Fetosense
A portable, wireless fetal heart monitor from CareNX that uses AI to read the trace, so non-specialist health workers can spot fetal distress and refer women in time.
- Finding unmapped children for vaccination
The IRC is piloting AI with satellite imagery to find children who never appeared in census records as conflict moves communities, and to adapt vaccination delivery routes as climate and conflict conditions change. The IRC estimates it could cut cost per…
- Mbaza
An AI chatbot that gives people reliable COVID-19 information such as case numbers, symptoms, vaccination, restrictions, test results and certificates. It works over basic phone messaging so no internet connection is needed.
- MERON
Kimetrica's MERON uses machine learning on a photo of a child's face to screen for severe acute malnutrition, as an alternative to arm-circumference or weight-for-height measurement; it reached proof of concept with UNICEF Venture Fund support.
- MEZA
WFP's MEZA digitises handwritten clinic records on child nutrition so programme officers far from the clinics can use the data.
- mimi-ml
PREDICTING RISK OF INADEQUATE MICRONUTRIENT INTAKE WITH TRANSFERABLE MACHINE LEARNING MODELS Identifying populations at risk of inadequate micronutrient intake is useful for governments and humanitarian organizations in low- and middle-income countries to…
- Médecins Sans Frontières (MSF)
An MSF Foundation project pairing a cheaper HPV test with AI that reads cervical images for precancerous lesions, to make cervical cancer screening workable in low-resource clinics; a study with 6,000 women ran in Malawi.
- Médecins Sans Frontières (MSF)
MSF ran a mobile X-ray truck in Manila's Tondo district where AI software scored chest X-rays for signs of tuberculosis, so only likely cases needed confirmatory testing; Epicentre studied whether it worked and was accepted.
- OTTAA Project
First AI based, predictive AAC platform for people with speech impairment
- PreSCo
Helping the diagnosis of infants affected by neonatal sepsis using machine learning
- PROMPTS with UlizaLlama
A two-way SMS service that sends pregnant women and new mothers stage-appropriate health messages and answers their questions through an AI-enabled help desk. A Swahili-speaking LLM sorts incoming messages into 30+ categories and flags urgent danger signs…
- Sahara (Intron Voice AI)
Speech recognition built for African accents and for speech that mixes African languages with English, first developed so clinicians could dictate notes into electronic medical records. It now also offers text-to-speech and voice agents used in health…
- Trauma Detection (LASSO)
A screening method for detecting trauma, built with the LASSO statistical technique. The authors describe it as rapid, low-cost and non-invasive, and note that it is specific to its context and needs retraining and validation elsewhere. Studied in Jordan,…
- VaccineLedger
VaccineLedger is a decentralized supply chain management platform that leverages blockchain technology, Internet of Things (IoT) and Artificial Technology and Machine Learning to capture data for every vaccine and other medical products in the supply…
- Wadhwani AI
Wadhwani AI is building AI tools for India's TB programme with the Central TB Division: reading line-probe assay results, predicting which patients will drop out of treatment, and screening by ultrasound.
- World Health Organization (WHO)
Two WHO mobile vans in Cox's Bazar, including the Rohingya camps, carry X-ray and molecular TB testing with AI-assisted reading of the X-rays, run with the Government of Bangladesh and the World Bank.
Information Management & Analysis 55
- ACLED
ACLED folds event data from its partner ExTrac, which tracks claims by Islamic State, into its Afghanistan conflict dataset (coverage from August 2019).
- AIDA (Artificial Intelligence for Development Analytics)
Search and question-answering tool over UNDP's archive of nearly 7,000 evaluation reports. It started in 2022 as an NLP prototype that pulled evidence out at paragraph level and now works as a chatbot that answers questions and cites the reports it draws on.
- AIDR
QCRI's AIDR uses machine learning to sift social media posts and images during disasters and pull out reports of injuries, needs and damage for responders; it was used after the 2015 Nepal earthquakes and the 2020 Beirut explosion.
- Ask ReliefWeb
A chat tool from ReliefWeb, OCHA's humanitarian information service, that answers questions from the reports in ReliefWeb's database, cutting search from hours to seconds. Built with AWS on Amazon Bedrock and announced in April 2025; whether it is open to…
- Beto
A conversational AI agent that carries out phone and voice impact surveys end to end, mixing closed and open-ended questions. It asks up to two follow-up probes when answers are unclear and checks incomplete responses in real time.
- Causal Map
A web app that uses a large language model to pull causal claims, with verbatim quotes, out of qualitative interview transcripts and then combines them into auditable causal maps. It is aimed at evaluators who need to synthesise what stakeholders say…
- Central Data Dictionary
IOM's DTM Central Data Dictionary is a bank of standard displacement-tracking indicators for building surveys, with AI-assisted search to find indicators.
- CODA
Africa's Voices built CODA, a tool where machine learning suggests codes for large volumes of SMS replies to interactive radio shows and analysts confirm them; first used on more than 250,000 messages with UNICEF Somalia.
- conflict-nlp
This project applies NLP techniques including topic modeling and large language models to analyze conflict and protest event data from the Armed Conflict Location & Event Data Project. Case studies focus on Iran, covering protest detection, conflict event…
- data-ai-chatbot
World Bank Data AI Chatbot: natural-language access to development data via Data360 MCP, charts and documents, and Proof-Carrying Numbers (PCN). Next.js + FastAPI. Forked from Vercel AI Chatbot. Apache-2.0 + World Bank IGO Rider.
- DEEP
The DEEP is a shared platform where humanitarian analysts upload, tag and analyse reports and other text, with AI suggesting tags, to support needs assessments; it is governed by a board of UN agencies and NGOs and built by Data Friendly Space.
- DEEP
WFP's in-house DEEP software analyses drone photographs of disaster areas with machine learning to map damage, cutting assessments from weeks to hours; first used after cyclones Idai and Kenneth in Mozambique in 2019.
- DevelopMetrics
DevelopMetrics sells language models tuned for international development work, used for knowledge management, evidence portals, project design, monitoring and document automation; USAID is a named client.
- DISHA
DISHA is a coalition led by UN Global Pulse that gives humanitarian agencies responsible access to data and ready-made AI models for damage assessment, socio-economic mapping and settlement mapping after disasters.
- Ecopia
Ecopia AI extracts buildings, roads and land cover from satellite imagery to produce detailed base maps, sold to NGOs and agencies for field planning and monitoring.
- EDDA
Software from Oikon and Fraunhofer ITWM that analyses drone images on site, without internet, to find destroyed buildings and estimate affected people after a disaster; it was to be donated to WFP and other responders.
- Evaluation Assistant
Internal generative AI platform that searches and summarises about 300 UNFPA evaluations and 1,400 related documents from the past 14 years. It can synthesise across country, regional and global levels, check an uploaded draft programme document against…
- fAIr
HOT's fAIr is an open-source tool that lets community mappers train their own AI models to detect buildings and other features in satellite imagery and speed up mapping in OpenStreetMap, keeping control of the model with local mappers.
- First Alert for Nonprofits
Dataminr's AI-driven real-time alerting on breaking crises and threats, offered free or discounted to nonprofits and UN agencies for situational awareness and staff safety; about 185 organisations use it.
- GeoPoll AI-Moderated Voice Surveys
An AI moderator conducts natural voice interviews with respondents on WhatsApp, asking questions, probing vague answers and transcribing and coding responses as they arrive. Results are delivered as audio files, structured data and live dashboards; a…
- HDX Signals
Monitors key humanitarian datasets (conflict, food insecurity, displacement, markets, agricultural hotspots) and sends alert emails when a situation gets significantly worse. Each alert includes a short situation summary generated by GPT-4o from public…
- High Resolution Population Density Maps
Meta's high-resolution population density maps combine satellite imagery with census figures to estimate where people live, published openly for more than 160 countries.
- HOT fAIr
An open-source AI system from the Humanitarian OpenStreetMap Team that rapidly maps intact buildings, roads and water infrastructure, building on community mapping. Referenced in the source for Syria; it is not shown as deployed there.
- Humanitarian Data Insights Project
DataKind, Save the Children and Microsoft are building a generative AI assistant and 'data recipes' that let humanitarian staff of any skill level ask questions of humanitarian datasets such as HDX in plain language; all outputs are to be open source.
- HumBERT
HumBERT is an open language model from The DEEP, adapted to humanitarian reports and news so it handles sector vocabulary better when classifying text.
- HumSet
HumSet is a dataset of humanitarian crisis documents from 2018 to 2021 in which experts marked key passages and tagged them by humanitarian analysis framework, for training text-classification models.
- HumVI
A Dataminr-led research dataset and models for spotting news reports of violence that affects aid work, in English, French and Arabic, sorted by sector such as food security, health and education, with annotations from Insecurity Insight.
- IMPACT Initiatives
IMPACT Initiatives used Meta Data for Good population and displacement estimates, built with AI models, to track where refugees from Ukraine were moving in near real time, shared with responders through UNHCR coordination meetings.
- IMPACT Initiatives
REACH worked with UADamage on an AI-assisted first pass to detect damaged homes in Nikopol from satellite imagery, then checked the results by eye.
- ImpactAI
Conversational agent that pulls together findings from thousands of impact evaluations and other rigorous causal studies. It answers questions about what works in development and shows sources and like-for-like comparisons across interventions.
- International Committee of the Red Cross (ICRC)
The ICRC used Sprinklr's AI social-listening dashboards in Afghanistan to track online conversations, spot misinformation and support staff safety and aid delivery.
- Interview GPT
An open-source web app that runs qualitative interviews as AI-led conversations, with participants answering by text or voice on any device. It analyses responses, summarises key findings after each interview and gives administrators a dashboard with…
- IOM (International Organization for Migration)
IOM's Displacement Tracking Matrix uses machine learning in its quality control: flagging unusual patterns in displacement data, classifying urban and rural sites, and analysing drone images of camps to judge conditions.
- iVerify
UNDP's iVerify helps national partners detect and answer false information, combining automated monitoring with human fact-checkers on the open-source Check platform; it is registered as a Digital Public Good.
- Meta
Meta's Population During Crisis maps compare how many Facebook users with location services are present in an area after a disaster against a normal baseline, to indicate displacement.
- Meta
Meta's Movement During Crisis maps show aggregated movement of Facebook users with location services between areas during a crisis, to help responders understand displacement.
- ML Enabler
Service integrating machine-learning model outputs (e.g. building or damage detection) into OpenStreetMap-based mapping applications, built for humanitarian mapping workflows with HOT.
- Nightlight analysis
Machine-learning analysis of night-light reflectance in satellite imagery, used to estimate damage and electricity provision. Documented in the Syria/ Turkey 2023 earthquake response.
- Norwegian Refugee Council (NRC)
NRC and the GIS company Alcis trained a deep-learning model to count tents in satellite images of displacement camps in Afghanistan, giving a fast estimate of how many people had arrived after the 2018 drought.
- OCHA
4W Wizard, built with SAP, uses machine learning to clean up and classify the 'who does what, where' reports OCHA receives from many partners, cutting the manual work of compiling them.
- OpenEval
AI-assisted search across Finland's published development evaluation reports. Users can query many reports at once and group results by theme, country or type of cooperation to get an overview of results and recommendations.
- Pomelo
Pomelo is a deep-learning method that turns coarse census counts into 100-metre population maps using building footprints and other open geodata, tested in seven sub-Saharan African countries; the code is public.
- PulseSatellite
UN Global Pulse and UNOSAT built PulseSatellite so UN analysts can run machine-learning models on satellite images, for refugee shelters, roof density and flooding, and correct the results to improve them.
- Radio Content Analysis Tool
Records talk radio across dozens of stations, uses automatic speech recognition to transcribe local-language conversations and flag clips on chosen themes, and passes them to human analysts who log insights in a searchable database. It was used to surface…
- RAMP
RAMP is an open-source deep-learning model that maps building footprints from satellite images in low- and middle-income countries, and local teams can retrain it for their own area without deep technical skills.
- segment-anything-services
Runs Meta's Segment Anything Model (image embedding, prompting and mask generation) as TorchServe microservices for geospatial imagery.
- SitRep Summary
Web app where a user uploads a batch of documents about one emergency and gets back a single draft situation report in UN OCHA style, with citations, key figures and a chart where the data allows. The user chooses the output language and reviews the draft…
- SKAI
SKAI, from WFP and Google Research, uses machine learning on satellite images to assess building damage after a disaster within hours instead of waiting for field teams; its code is open source.
- Sophia
ACAPS' Sophia is an AI platform that helps ACAPS analysts collect, translate and process large volumes of crisis information quickly, including for its severity indices.
- Sudan Horizon Scanner
A UNDP system for Sudan that uses AI to map development interventions, run programme simulations, pick up early warnings and analyse public conversation, to support decisions.
- UNDP RAPIDA and Crisis Response Hub
UNDP's platform combining satellite imagery, social media signals and night-light data for crisis analysis. It includes EVA.ai workforce matching, which the source cautions may favour profiles with more digitally visible experience over local expertise.
- UNHCR
A UNHCR project in Hungary using generative AI to follow refugees' online conversations, spot emerging needs and threats, and improve online engagement.
- Ushahidi
Through Dataminr's AI for Good programme, Ushahidi received three custom models that sort citizen reports into categories, improve their location and handle several languages, replacing slow manual processing of crowdsourced election data.
- World Food Programme (WFP)
WFP has run a humanitarian drone programme since 2017: prepositioned equipment in five regions, emergency mapping and training for responders in 15 countries, with working groups on regulation, ethics and imagery analysis.
- xView2
A machine-learning model that segments satellite imagery to identify damaged buildings, reported at 85 to 90% accuracy. Used by the UN, World Bank, IFRC and WFP in the 2023 Türkiye-Syria earthquake response, not in routine Syria programming.
Logistics & Supply Chain 9
- AHEAD
WFP's AHEAD project tests remotely driven, partly autonomous all-terrain trucks, including an amphibious version, to deliver aid where it is too dangerous for drivers; tested in Germany with field trials planned in Africa.
- Camp Forecast
Camp Forecast is described as an AI demand forecasting tool for refugee and IDP camps. It predicts the number and type of residents and the supplies they will need over short, medium and long terms, to help camp managers plan.
- ForLab Plus
ForLab Plus is a forecasting and procurement platform through which Ethiopian public health facilities submit consumption-based forecasts and procurement commitments. It supports facility-level forecasting of health commodity needs to guide national…
- Optimus
Optimus is a WFP decision tool that works out the cheapest nutritious food baskets and delivery routes for an operation, from population, route and nutrition data; it has been used in 44 WFP operations.
- SCOUT
SCOUT is an AI optimisation tool for planning WFP's food supply chain upstream of recipient countries. Using prices, transport costs, lead times and access constraints, it recommends where to buy food, where to store it and how to deliver it.
- STRIATA
STRIATA uses machine learning to predict demand for health goods and services at each facility: who will turn up for care, what each facility will use, and each facility's capacity. Its origins are a predictive vaccine supply chain that forecast…
- UN AI Smart Mission Planner
An AI tool, built by WFP with Google, that plans humanitarian missions by optimising flights, vehicles and guesthouse bookings across UN agencies through the UN Booking Hub.
- UNICEF
Pendulum, an AI supply-chain company, describes work on machine-learning demand prediction to improve immunisation coverage.
- UNICEF
UNICEF's Copenhagen supply warehouse uses eight computer-controlled robot cranes to store and retrieve some 36,000 pallets, speeding up emergency dispatch.
Other 3
- extractor-operational-learnings
IFRC operational reports analysis using Natural Language Processing
- higgs-tts-3-4b
Multilingual text-to-speech model with expressive, controllable speech, described as suited to voice agents.
- OmniTranslate-1.1-GGUF
Supports 500 languages.
Productivity & Internal Operations 37
- AI Sandbox
WFP's AI Sandbox is an internal space where teams test AI ideas safely, with expert support, before scaling them; its 2024 beta covered eight uses including fraud detection, evidence retrieval and chatbots.
- AIforNGOs
A platform for Indian nonprofits that provides AI-assisted proposal and report templates, donor research and funding-landscape analysis, and automation of social media and email campaigns. It is paired with a funding-alert service and training.
- ALMA
An AI fundraising agent for nonprofits that runs personalised conversations with prospective and lapsed donors over WhatsApp and email, aiming to get them to register, donate, subscribe or resume giving. Staff can review and approve generated messages or…
- AnswerMate
Internal staff chatbot that answers questions about ITCILO rules and procedures. It draws on governance documents such as the Statute, Regulations, Rules and Circulars, and links each answer to the source PDF so staff can check it.
- Azure OpenAIText Classification Service
An internal OCHA service built on Microsoft's Azure OpenAI that sorts text such as pooled-fund project summaries and incident reports into categories automatically, saving staff from reading and tagging them by hand.
- BEYOND
Monitors UNHCR's private-sector donors and partners between annual due-diligence screenings. It takes partner records from UNHCR's Synergy (Salesforce) platform, scans news and ESG databases using multi-language keyword monitoring, and alerts partnership…
- British Red Cross
A four-month British Red Cross proof of value that used software robots (Blue Prism robotic process automation) to check supplier invoices, freeing finance staff for other work, and looked at other HR and IT processes to automate.
- Candid fundraising assistant
An assistant inside Candid Search that helps nonprofits find funders whose giving matches their mission and location, check eligibility, see which funders back similar organisations, size a request and draft a letter of intent, drawing on Candid's grants…
- Capability Coach
Planned chatbot on the IFRC Red Solutions Hub. It will answer National Society staff questions using the Hub's knowledge repository and point them to relevant capability solutions, resources, examples and community posts.
- Charity Excellence AI Bunny
A free AI assistant inside the Charity Excellence platform that helps small charities draft funding bids and find funders, built on ChatGPT.
- CharityGPT
An AI feature in the Botim messaging app that lets users make charitable donations through a conversational interface. Launched for Ramadan, it lets NGOs set up fundraising campaigns via Astra Tech and Ma'an.
- Dataro
Predictive fundraising software that connects to a charity's donor database (Salesforce, Blackbaud, Dynamics and others) and ranks supporters by how likely they are to give, become regular donors or leave a gift in their will, so appeals can be targeted.…
- docs-rag
RAG LLM model and chat interface, providing a user-facing knowledge base of HOT's tools
- Donor Retention AI assistant (AI for Impact 2024)
A project to reduce churn among UNICEF monthly donors. Churn-prediction models (XGBoost, random forest and others) performed poorly, so the team switched to a chat assistant that turns plain-language questions into SQL queries on donor data and explains…
- Dr.Cai
Dr.Cai is a chatbot on the Danish Refugee Council's careers pages that answers questions about working at DRC and applying for jobs.
- eLUNa with TAPTA4UN
In-house, browser-based translation tool for UN translators. It reuses earlier translated sentences and UNTERM terminology and links to UN documents, and offers machine translation (TAPTA4UN, trained on UN documents) for new sentences.
- Financial Report Reviewer
An on-device AI review tool that screens partner financial reports against donor rules before funds are released or reports are consolidated. A reviewer uploads a partner's transaction list and budget; the tool runs 30+ checks across six areas (reporting…
- Fundraising Digital Teammate
An agentic AI fundraising assistant that identifies potential donors, matches funders to an NGO's strengths and helps build a fundraising plan. It also produces strategy outputs such as SWOT analyses and donor landscape maps that small NGOs usually buy…
- Grant Assistant
A proposal-writing platform that summarises requests for proposals, suggests structure from templates, generates draft content and runs a compliance review before submission. It also matches organisations to funding opportunities using semantic search.
- Grant Wizard
An AI platform from Nigeria's Grant Master that helps African nonprofits, startups and social enterprises find and apply for grants: it matches an organisation's profile to opportunities by sector, geography and eligibility, analyses funder requirements…
- Grantable
A grant-writing platform that finds aligned funders with a plain-language explanation of each match, drafts proposals section by section from an organisation's own past proposals and documents, and tracks deadlines and the grant pipeline. Has a free plan;…
- Grantora
Helps small and mid-sized NGOs prepare grant proposals that follow donor-expected sections, terminology and compliance, drafting in the organisation's working language and exporting a structured draft for internal review. It does not submit proposals or…
- GrantWise AI
A planned premium AI layer for GrantsDatabase, an African funding-opportunity platform, intended to match users with relevant grants, fellowships and calls. The platform offers curated discovery, eligibility and fit assessment and grant advisory support.
- Instrumentl
A subscription platform that uses AI to help nonprofits find grants that fit them, research funders and manage applications and deadlines from discovery to reporting. Raised $55 million in 2025 to expand its AI features.
- Kindsight AI agents
AI agents for fundraising teams launched in September 2026: one plans donor visits for a trip and drafts outreach and briefings, another turns plain-language questions into wealth-screening runs, and a third writes two-page prospect briefs from giving…
- Knowledge chatbot
An internal chatbot that lets NRC programme staff ask questions about the organisation's policies and guidance in plain language, answering from NRC's own documents with links to the source. It runs a small Llama model in NRC's own cloud to keep data…
- Kobo, Inc
KoboToolbox added automatic transcription and machine translation of recorded interviews, so humanitarian teams can analyse what affected people say across languages; transcription covers 72 languages and translation 106.
- MALENA
Users upload a document and the tool picks out environmental, social and governance (ESG) information, such as greenhouse gas emissions and gender data, and scores the sentiment around each term. It was trained on more than 15 years of IFC emerging-market…
- Mercy Corps
Mercy Corps published documentation for a proof-of-concept chatbot on Microsoft Azure OpenAI that answers staff questions only from its own document library, as a safer alternative to public chatbots.
- Methods Matcher
Methods Matcher is Mercy Corps' assistant for its own aid workers: it searches past projects, research and proven practice and returns short, evidence-based answers for programme decisions. Launched in November 2024 with Cloudera and used across dozens of…
- NGOwise.ai
An early-stage platform for small and grassroots civil society organisations that combines funding search, help drafting proposals, project tracking and donor reporting in one place, aimed at groups without dedicated fundraising or compliance staff.…
- Plinth AI Grant Writer
Searches a grant database from a plain-language project description, compares each funder's criteria against the charity's profile to flag matches and gaps, and drafts application answers from the charity's own past applications, reports and website using…
- Propel
Propel is a platform where development and humanitarian teams record lessons from their projects; AI tags each story and suggests similar lessons from elsewhere in the organisation.
- Ukrainian Red Cross
VRP Consulting set up Salesforce Nonprofit Cloud and Marketing Cloud for the Ukrainian Red Cross, pro bono in five weeks, to handle the surge of donations in 2022 and track donors.
- UNHCR
A UNHCR project in Iraq testing language-processing tools to translate, summarise, sort and transcribe the text and voice feedback that communities send in.
- UNifyHR
A shared chatbot that answers HR policy questions for 13 UN organisations, including UNHCR, UNICEF, WFP, IOM and UNDP, and lets HR teams compare policies across agencies. Conceived by UNHCR and built by UNICC; launched in October 2024.
- Veda AI
A fundraising platform for Indian nonprofits combining a CSR-company and donor database with fit-scored fundraising strategy recommendations. Its Veda AI drafts grant proposals from an organisation's knowledge hub and an RFP, and AI agents enrich and score…
Protection 17
- Adala-Bot
An Arabic-language chatbot launched in March 2022 by Med Dialogue for Rights and Equality, available around the clock. It stores no personal data and deletes conversations within 24 hours.
- Arabic Violence Indicator (and Hebrew Violence Indicator)
Language-model classifiers that automatically monitor social media for hate speech and incitement in Arabic (and a separate Hebrew version) in the Palestinian context. Results, including reach and engagement, are published in near real time on 7amleh's 7or…
- CRC Automatic Classification
Machine-learning project classifying Committee on the Rights of the Child (CRC) recommendations into thematic clusters, bottlenecks and sub-domains of the ECA Child Rights Monitoring Framework, to speed up tracking of how governments respond to CRC…
- CRC-Automatic-Classification
A project leveraging machine learning to classify CRC recommendations into thematic clusters, bottlenecks, and sub-domains of the ECA Child Rights Monitoring Framework. This project enhances efficiency and accuracy in monitoring child rights, supporting…
- Dataminr
UN Human Rights and Dataminr trained a model that finds and classifies reports of attacks on human rights defenders in online news, and handed over the data, training scripts and instructions so it can be rebuilt.
- Generation AI
Generation AI is a UNICEF research and policy initiative, with partners including the World Economic Forum and UC Berkeley, to make sure AI is designed with children's rights in mind.
- Girls' Rights Platform database
HURIDOCS, Plan International and Google.org fellows added machine learning to the Girls' Rights Platform so new policy documents are imported and tagged by topic and affected group automatically, leaving staff more time for advocacy.
- International Committee of the Red Cross (ICRC)
An ICRC proof of concept that uses language models to re-sort ACLED conflict records by international humanitarian law categories and pull out who did what to whom, tested on Colombia, Ethiopia and Ukraine.
- LexAid
A tool built by SidLabs with AsyLex that drafts legally grounded asylum appeals in Switzerland, which lawyers then review; a 2025 AI for Good award finalist.
- OlimpIA
A WhatsApp chatbot that gives people facing digital sexual violence legal guidance and emotional support, including tools for anxiety and panic attacks. Separate AI models act as lawyer, digital defender and psychologist, while others act as safety filters…
- Online Grooming Detection Classifier
An eight-week Omdena challenge with Save the Children that built machine-learning classifiers, a labelled dataset and a browser extension to detect grooming in online chats.
- Phoenix
An open-source process and toolkit that helps peacebuilders and mediators gather social media data, classify it with customisable machine learning models and explore it in dashboards. It is used to understand online conflict dynamics and design…
- RiMAP AI co-pilot
A generative AI co-pilot that helps UNHCR protection staff research and analyse refugee law and policy, built on UNHCR's rights-mapping platform.
- Safe Pro AI SpotlightAI
Detects over 150 types of landmines and unexploded ordnance with over 80% accuracy, in a deployment in Ukraine with HALO Trust, NPA and UNDP. The source states it has not yet been fully deployed in Syria.
- Save the Children
An Omdena challenge with Save the Children that mapped Kenya's child protection system and built predictive models and recommendations to strengthen reporting and response to child labour, trafficking and abuse.
- Sophia
A free, round-the-clock chatbot from the Swiss charity Spring ACT that helps survivors of domestic violence record abuse, learn their rights and find local help, without leaving a trace on their phone.
- UNHCR
A UNHCR-funded project in the Republic of Korea using AI to draft court submissions that challenge refusals of refugee status, to widen refugees' access to justice.
Resettlement & Integration 2
- Annie MOORE / Annie 2.0
Software that uses machine learning and optimisation to match resettled refugees with host communities, weighing refugees' needs against local capacity; in use with a US resettlement agency since 2018.
- UNHCR
UNHCR is using AI to analyse how refugee status determination cases move through its processes, to find where delays build up and ease backlogs.
WASH 2
- IWMI (International Water Management Institute)
An IWMI project under the CGIAR AI Hub building an AI framework on global datasets to produce basin-scale water models quickly, especially where local data are thin; first used with the Limpopo Watercourse Commission.
- SWOT
A free, low-bandwidth, web-accessible tool that sets site-specific chlorination targets for water safety, developed by York University's Dahdaleh Institute with MSF. It assumes consistent, high-quality routine data collection and requires adequate training…