Agriculture, the backbone of the Indian economy, provides livelihoods to over 40% of the population. Yet, estimates of national agricultural production and macro-level decisions on crop planning, exports, and government schemes often also rely on data created at desks, alongside scientifically collected data and limited on-ground evidence. We observe that end-user stakeholders often struggle to participate in decision-making. Thus, a divide persists between data collection methods, policymakers, and on-ground lived experiences. To facilitate communication of ground-level information, we propose Participatory Insight, a design research method for grounded, nuanced and potentially scalable stakeholder engagement in policymaking. We instantiate this method through Avani, an AI-supported, audio-based data collection and analysis system designed for the agricultural domain. Avani aims to deliver actionable insights to decision-makers to reduce information gaps. Our main contributions include a generalisable method for participatory policymaking, and a pilot testing its implementation in the context of Indian agriculture.

Web-augmented large language models (LLMs) promise end-to-end retrieval agents for time-sensitive information access, yet their reliability in low-resource environments remains unclear. We introduce MarathiWeb, a benchmark for evaluating LLM-native web retrieval in Marathi that separates translated global queries from natively authored, local information needs. We evaluate two frontier LLMs with built-in web search against a controlled Google-based retrieval baseline, and analyse retrieval depth, failure modes, and cost-accuracy trade-offs. Despite high tool invocation rates, accuracy drops sharply in Marathi, most severely for local queries, with the dominant failures arising from evidence integration rather than missing sources. We further find that shallow retrieval (one web call) is consistently the most accurate and cost-efficient strategy, while deeper multi-step search degrades performance. MarathiWeb exposes persistent gaps in local-language information access and provides an open benchmark for low-resource retrieval research.