Managing groundwater availability is critical everywhere, but especially in India where nearly 25% of global groundwater is used. Depletion of critical aquifers threatens states like Punjab, Rajasthan, and Andhra Pradesh, where groundwater levels have plummeted below sustainable thresholds. This crisis endangers food security, rural livelihoods, and water sustainability for the whole nation.

The “Wise Drop” solution uses agentic AI to provide user-specific experiences, empowering everyone involved in managing groundwater with data and insights specific to their role. Farmers can interface simply through local-language Whatsapp and get personalized irrigation advice and crop recommendations to help them reduce water usage and improve yields. Local water managers can monitor aquifer health and contamination risks, and get reports with prioritized actions, allowing them to optimize resources more effectively. Policy-makers can run simulations on the impacts of different intervention methods and generate data-driven policy recommendations that consider agricultural productivity and aquifer preservation.

The solution leverages IBM watsonx.ai and Granite series models (Granite-3-2-8b-instruct, Granite-guardian-3-8b, Granite-13b-instruct-v2, Granite-20B-multilingual) within a multi-agent workflow that integrates historical data, real-time weather inputs, and contamination reports. Each specialized agent uses a specific Granite model best suited for its function.

In moments of urban crisis such as infrastructure failures, flooding, fires, or road accidents, delays in coordinating the appropriate response can be critical. In rapidly growing urban areas like Bangalore, a few lost minutes can cost lives and property. Current public emergency systems rely heavily on manual reporting, which makes it challenging to assess, prioritize, and route help to incidents in real time.

The “Alert X” solution from team Agents FDRK seeks to make cities safer, more connected, and resilient. It helps reduces the gap between incident detection and emergency action from minutes to seconds. It’s is an autonomous, multi-agent AI framework that detects, analyzes, and communicates critical incidents to authorities with intelligent coordination. Built on IBM watsonx.ai and IBM watsonx Orchestrate, Alert X integrates visual understanding, contextual reasoning, and natural interaction into one continuous workflow. One agent captures live video frames from a mobile camera and uses Granite Vision or Llama-based visual models to identify the scene (e.g., flooded streets, fire, damaged infrastructure). It then produces a natural-language summary with detected risks and objects. Another agent uses Openstreet Map APIs to retrieve information on nearby critical facilities such as hospitals, schools, crowded spots, connectivity hubs, police stations, fire stations, and shelters. A third agent uses Granite 3.3 Reasoning LLM to analyze both the visual and the geographical data. It evaluates severity, impact, and urgency, then recommends actionable steps such as “Evacuate nearby area,” “Ensure no children in the surrounding schools go outside,” “Notify fire department,” or “Divert vehicles to alternate route.” Finally, after user approval, a fourth agent generates a human-like conversational script to brief authorities. It mimics a real person’s tone providing event context, urgency, and help needed, ensuring clear, empathetic communication.

Watsonx Orchestrate runtime triggers each agent, monitors outputs, and maintains the contextual flow. The team hopes to implement more agents in the future such as a drone data assessor or citizen chatbot.