PI: Milind Bunyan
Team: Seena Narayanan Karimbumkara, Smrithi Gnaneshwar, Srijana Sharma, Aprajita Jayaram, Siddappa Setty, Preeti Singh, Nakul Chettri, Vijay Rasquinha, Sujata Seshan
Overview
As part of ATREE’s and the Biodiversity Collaborative’s efforts to make scientific, technical, and locally relevant knowledge more accessible to communities, site-specific AI-enabled WhatsApp chatbots are being developed for two biodiversity-rich landscapes: GramaMitra for the Biligiri Rangaswamy Temple (BRT) Tiger Reserve landscape in Karnataka, and GaonSahayog for the Sikkim–Darjeeling landscape in the Eastern Himalayas.
The chatbots are being developed as landscape-specific knowledge systems that bring together scientifically curated resources, expert knowledge, government information, and community-generated questions. Rather than functioning as generic information platforms, their content is tailored to local crops, livelihoods, ecological contexts, languages, and community information needs.
In the last year, we progressed from knowledge-base development and pilot testing to publicly deploying GramaMitra in BRT, while GaonSahayog advanced through its first pilot and preparations for a second round of community testing in the Eastern Himalayas.
GramaMitra – BRT Landscape
GramaMitra was launched in the BRT landscape on 17 June 2026, marking the transition from development and testing to active community-facing deployment. The launch brought together Soliga community members, Farmer Producer Organisation representatives, school teachers, researchers, and ATREE & Adavi staff. The programme was also attended by Mr Sameer Shisodia, Nayantara Lakshman, Siddharth Lahri and Sailee Rane from the Rainmatter Foundation; Dr Ravi Chellam (CEO, Metastring Foundation & Coordinator, Biodiversity Collaborative); and Ms Anita Arjundas (Executive Director, ATREE), whose interactions with community members and the project team added valuable perspectives to the discussions on the chatbot use case.
GramaMitra currently includes curated information on coffee, honey, and pepper, with planned expansion to select non-timber forest products (NTFPs), along with information related to government schemes, biodiversity, and human–wildlife interactions. The content has been developed through a combination of literature and resource compilation, expert consultations, field interactions, and community-generated questions.
Before launch, the coffee and honey knowledge bases underwent extensive refinement. Expert interviews were completed, responses were transcribed into FAQs and resource sheets, and knowledge gaps identified during field testing were addressed. Kannada translations were incorporated to improve accessibility and response quality, supported by a glossary of technical and locally relevant terminology.
Pepper was subsequently added as a new thematic area, with a resource master sheet, FAQs, and Golden Questions prepared. Work has also begun on the next phase of content development, with NTFPs as a priority area.
The launch gave community members an opportunity to interact directly with GramaMitra and explore its utility. Initial interactions and subsequent feedback helped identify areas for further improvement, particularly regarding Kannada translation, response accuracy, and gaps in dynamic or frequently changing information.
The project team conducted a post-launch review to consolidate these learnings and identify priorities for the next phase. Outreach activities are now being planned to increase awareness and encourage wider and sustained adoption of GramaMitra among community members and local civil society organisations (CSOs).
GaonSahayog – Sikkim–Darjeeling Landscape
GaonSahayog is being developed to support communities and agricultural stakeholders in the Sikkim–Darjeeling landscape. This chatbot focuses on locally important crops and livelihood concerns, with the initial thematic areas including orange, avocado, large cardamom, and red cherry pepper (commonly called dalle).
Here too, the knowledge base is being developed using research literature, expert consultations, government and institutional resources, and community-generated questions. Particular attention has been given to large cardamom, an important livelihood crop in Sikkim, with information being compiled on cultivation practices, planting material, crop management, pest and disease management, harvesting, post-harvest handling, climate-related challenges, and relevant support programmes.
A first community pilot of GaonSahayog was conducted in the last week of May 2026. The pilot provided an opportunity to test the chatbot with intended users, understand how communities interact with the platform, identify gaps in the knowledge base, and assess language, navigation, and response-related issues.
Since the first pilot, the team has been refining the knowledge resources, FAQs, Golden Questions, and chatbot responses based on the feedback received. Nepali translations are also being incorporated to improve accessibility for Nepali-speaking users.
A second pilot of GaonSahayog is planned for the third week of September 2026. This round of testing will assess the improvements made following the first pilot and provide further insights into user experience, content relevance, language accessibility, and the information needs of agricultural communities. Feedback from this pilot will be used to further refine the chatbot before the launch of GaonSahyog.
Community Engagement and Learning
Community engagement is an integral component of the chatbot development process. The project follows an iterative cycle of content curation → expert validation → pilot testing → community feedback → refinement → further testing.
This approach ensures that the chatbots are not developed solely as repositories of scientific information but evolve in response to actual community information needs. Questions raised by users during pilots and outreach activities are being documented and used to identify gaps and develop additional FAQs and knowledge resources.
The project also recognises the importance of combining scientific knowledge with locally held and traditional ecological knowledge. By providing a platform for communities to ask questions, provide feedback, and identify information gaps, the chatbots can support an enhanced two-way relationship between researchers and communities.
Technical and Institutional Partnerships
- Tech4Dev has provided core technical support for chatbot development, including Glific integration and facilitating the technical connections with Meta and Gupshup required for WhatsApp-based chatbot deployment. They have also supported integrating a live weather API to enable real-time weather information through the GramaMitra chatbot, and supported initial AI evaluations to assess and improve chatbot performance.
- A partnership with Haqdarshak is being explored to integrate government-scheme APIs, enabling communities in both landscapes to access relevant scheme information and application details directly through the chatbots. The partnership will also support the development of a more dynamic and regularly updated government-scheme information system, strengthening the practical utility of GramaMitra and GaonSahayog for the community.
Monitoring, Evaluation and Learning
A Monitoring, Evaluation and Learning (MEL) framework will assess the reach, adoption, use, and usefulness of the chatbots. Baseline information will help establish the existing sources of information communities use for agriculture, government schemes, weather, markets, forest-based livelihoods, and human–wildlife interactions.
Following deployment and wider outreach, usage patterns, user feedback, community interactions, and endline assessments will help determine whether the chatbots improve access to reliable, locally relevant information. The MEL process will also generate learning to guide future development and create a blueprint for site-specific digital knowledge platforms.
Key Achievements to Date
- Developed two landscape-specific AI-enabled WhatsApp chatbots: GramaMitra and GaonSahayog.
- Launched GramaMitra in the BRT landscape on 17 June 2026.
- Conducted community-facing demonstrations and initiated user adoption.
- Developed curated knowledge resources, FAQs, Golden Questions, and expert-validated content.
- Developed Kannada-language content for GramaMitra and Nepali-language content for GaonSahayog.
- Initiated development of a structured MEL framework for assessing chatbot adoption and impact.
- Conducted the first GaonSahayog pilot and incorporated user feedback into subsequent development.
- Established institutional and expert linkages for strengthening technical content.
- Identified content gaps related to translation, response accuracy, dynamic information, and user needs.
- Initiated development of outreach and training activities to support wider community adoption.
Next Steps
The immediate priorities for the coming period include:
- Conducting the second pilot of GaonSahayog in the third week of September 2026 and incorporating findings into the chatbot.
- Continuing refinement of GaonSahayog’s agricultural knowledge base and expanding expert-validated content.
- The GaonSahayog chatbot is expected to be launched by the end of November 2026, subject to the completion of content development, testing, and validation.
- Expanding GramaMitra’s NTFP knowledge base and developing additional thematic content based on community needs.
- Planning to conduct outreach workshops and training sessions with communities and CSOs to increase awareness and adoption of GramaMitra.
- Continuing to improve Kannada and Nepali language accessibility and chatbot response quality.
- Strengthening the MEL framework and monitoring chatbot adoption, usage, and user experience.
- Using community-generated questions and feedback to continuously update and improve both knowledge bases.
Overall, the project has progressed from the initial development of site-specific knowledge systems to community-facing testing and deployment. The launch of GramaMitra in BRT and the two-stage pilot process for GaonSahayog in Sikkim–Darjeeling provide important opportunities to understand how AI-enabled tools can support community access to scientific and locally relevant knowledge. The learning from these two landscapes will inform the continued development of ATREE’s and the Biodiversity Collaborative’s broader approach to using digital technologies for sustainable livelihoods, community resilience, and biodiversity conservation.