#68: Impact Signals #68 — Mercy Corps Ships VERA Agentic AI, Google's Flood Forecasts Reach 2 Billion, IIT Mandi's Himalayan Landslide Warnings
AI for Impact Daily Briefing, July 09, 2026
Top Stories
Mercy Corps and Cloudera Ship VERA, an Agentic AI Crisis Analyst With Hard Savings Numbers
Cloudera and Mercy Corps announced VERA (Verified Evidence & Research Assistant) on July 9, an agentic AI solution that automates research, aggregates sources, and produces localized crisis-specific analysis for field teams, running on Cloudera AI Studios and powered by Anthropic's Claude models. The release caps nearly 2.5 years of collaboration and comes with rare production metrics from live operations: up to 90 percent time savings on election security reports in Colombia (about $2,000 saved per report), Sudan food security analysis cut from 5-6 days to 2-3 days (about $1,500 saved per report), and active use in disease outbreak monitoring across Central and East Africa. Mercy Corps operates in 46 countries and reached 36.5 million people in 2025. Why it matters: This is one of the first humanitarian agentic-AI deployments to publish per-report dollar and time savings from live operations rather than a pilot. NGO analytics and MEL teams should lift these report-level benchmarks to build their own business case, and the underlying pattern, an agentic research assistant over a curated evidence base, is replicable.
Google Puts Numbers on Its Crisis AI: Flood Forecasts for 2 Billion People, FireSat in Orbit, Flash-Flood Tools Open-Sourced
Timed to the AI for Good Summit, Google published a consolidated account of its crisis resilience stack on July 7 alongside a UN-aligned report on AI-enhanced multi-hazard early warning. AI river flood forecasting now reaches 2 billion people across 150+ countries with up to 7-day lead times, wildfire boundary tracking runs in 34 countries, and Public Alerts covers 90+ countries. New pieces include the first three FireSat wildfire-detection satellites launched with Earth Fire Alliance and Muon Space, the Google Earth AI collection of climate and geospatial models, and open-sourced flash-flood resources: the Groundsource dataset and a hydrology modeling framework. The DISHA damage-assessment workflow built with UN OCHA and UNOSAT has now deployed 11 times, including assessing 385,000 buildings after Hurricane Melissa. Why it matters: Emergency managers and humanitarian data teams can plug in today: Flood Hub likely already covers their operating areas, and the newly open-sourced flash-flood dataset and hydrology framework give national hydromet services and research NGOs a free starting point for local models instead of building from zero.
IIT Mandi Switches On Landslide Early Warnings for the Entire Indian Himalayan Region, With WhatsApp Alerts
Researchers at IIT Mandi, led by Prof. Dericks Praise Shukla, announced on July 8 a fully operational Landslide Early Warning System forecasting landslide probability across the whole Indian Himalayan Region, built on ensemble machine learning models trained on roughly 26,000 historical landslides from the Geological Survey of India database and combined with real-time rainfall data. Unlike India's earlier region-locked pilots, the system issues location-specific forecasts nationwide, publishes downloadable PDF bulletins, and lets users subscribe to WhatsApp alerts for chosen locations. It lands mid-monsoon, in the exact weeks Himalayan states see their deadliest slope failures. Why it matters: Delivery over WhatsApp is the actionable detail: district disaster management authorities, mountain NGOs, and road and trekking operators in Himachal, Uttarakhand, and the Northeast can subscribe to point forecasts now at zero cost and wire bulletins into existing community warning chains. It is also a template for other data-rich, infrastructure-poor mountain regions.
IRC Challenges the Tech Industry at AI for Good: Put AI to Work for 118 Million Displaced People
As its delegation joined the AI for Good Global Summit in Geneva on July 7, the International Rescue Committee publicly urged tech companies and philanthropies to fund AI tools that connect displaced people to healthcare, education, and trusted information, citing the new global figure of nearly 118 million forcibly displaced. The IRC backed the ask with its own deployments: Signpost, co-founded with Mercy Corps, has delivered verified, localized information to more than 20 million people in 30 countries since 2015, and Signpost and aprendIA are running purpose-built AI agents in Bangladesh and Nigeria helping displaced people access food, education, and legal documentation, drawing on models from OpenAI, Anthropic, and Google. Why it matters: This is a funding signal and a partnership door. Refugee-serving organizations can point funders at the IRC's framing, AI for client-facing information access rather than back office, and the Signpost playbook of verified local content plus AI distribution is openly documented and already multi-country, making it the reference architecture to copy or join.
Pakistan's National Emergency Center Enters Monsoon Season With AI Projections and Relief Drones
Pakistan's National Emergency Operations Center, run by the National Disaster Management Authority in Islamabad, is using satellite imagery and AI projections to monitor this year's erratic, El Nino-shaped monsoon, which officials warn could bring simultaneous floods and droughts across different regions. Monsoon rains have already killed 17 people and injured 41 nationwide since June 26. Officials say drones will fly surveillance over threatened areas and drop essential supplies to families cut off if disaster strikes, folding aerial logistics directly into the national response plan rather than treating it as a pilot. Why it matters: Pakistan was the site of the catastrophic 2022 floods that affected 33 million people, so its national adoption of AI-driven monsoon monitoring is a bellwether for climate-vulnerable states. Humanitarian teams operating in-country should sync early-warning thresholds and drone corridor plans with NEOC now, before peak monsoon, and other national disaster management authorities can study the NEOC setup as a South-South reference.
AI Forecast Gave Nashik a 24-Hour Head Start on Cloudburst-Like Rain
An AI-assisted weather forecasting system, reported as BharatFS running advanced atmospheric modelling on supercomputing, predicted cloudburst-like rainfall over Nashik, Maharashtra nearly 24 hours in advance. With the IMD forecasting up to 300 mm of rain, Chief Minister Devendra Fadnavis issued a public cloudburst warning for July 7 and authorities moved emergency measures into place ahead of the event, an early concrete case of India's new-generation AI weather models shifting a district-level response from reacting to rainfall to pre-positioning for it. Why it matters: A verified 24-hour lead time on hyper-local extreme rain is the difference between evacuation and rescue. State disaster authorities and city emergency planners across South Asia should track BharatFS-style localized AI forecasts as they become available and pre-agree trigger thresholds, which rainfall forecast activates which action, so the extra lead time converts into moves on the ground.
Upcoming Events & Opportunities
AI for Good Global Summit 2026 (ITU)
- July 7-10, 2026 (happening now)
- Location: Palexpo, Geneva, Switzerland
- 300+ sessions; this week's Mercy Corps, Google, and IRC stories all broke around this summit
- Register: aiforgood.itu.int
World Summit on AI
- June 2027, exact dates to be announced (successor summit announced July 8 at AI for Good)
- Location: Geneva, Switzerland
- Registration not yet open; watch aiforgood.itu.int
Anthropic Claude Corps Fellowship (Funding)
- Amount: $85,000 salary plus benefits for a 1-year, full-time nonprofit placement; $150M total program, 1,000 fellows planned
- Deadline: July 17, 2026 (first cohort of 100, starting October 2026; rolling for January and August 2027 cohorts)
- Eligibility: US work authorization, age 18+, under two years full-time work experience, any educational background
- Apply: anthropic.com
British Council Cultural Protection Fund (Funding)
- Amount: Grants up to £500,000 (about 7 awards at £100k-£500k plus about 20 awards under £100k)
- Deadline: July 20, 2026, 23:59 BST (Expression of Interest; full applications by August 28, 2026)
- Eligibility: Organizations protecting cultural heritage at risk in 17 countries including Sudan, Syria, Yemen, Nepal, and Pakistan
- Apply: cultural-protection-fund.britishcouncil.org
Active Disaster Monitoring (GDACS/OCHA)
- Northwest Pacific (Super Typhoon Bavi):** GDACS RED alert; maximum winds 287 km/h, an estimated 38.5 million people exposed to Category 1 or higher winds across Guam, the Northern Marianas, Japan, Taiwan, and China; associated storms and tornadoes in Hubei have killed at least 15, with national rescue mobilization ordered in China. Source: gdacs.org.
- China (southern/central floods):** GDACS Orange alert; 12 deaths and 85,197 displaced to date, compounding with Typhoon Bavi's rainfall. Ongoing since June 6.
- France:** Forest fires, GDACS Orange alert; suppression ongoing since July 4.
- Portugal:** Forest fires, GDACS Orange alert; suppression ongoing since July 2.
- Madagascar:** Drought, GDACS Orange alert; slow-onset food security impact in the Grand Sud, ongoing since April.
- Note: only major or ongoing-major disasters are featured; low-severity GDACS Green alerts (including M6.2 Halmahera and M6.1 Miyakojima earthquakes with green shaking alerts) are excluded per the major-only bar.
Sources: See individual stories above for full attribution.