evidoria

← Back to browse

Good practice Imported

Jugalbandi — A Multilingual Generative-AI Chatbot Connecting Rural Indians to Government Welfare Schemes

India · Nuh · See the India profile

Evidence: Descriptive / self-reported Top 75% 47/100 · Ask Evidence Copilot about this practice

AI4Bharat, OpenNyAI and Microsoft Research built Jugalbandi, a WhatsApp/Telegram chatbot using Bhashini speech AI and GPT-4 to explain scheme eligibility in 10 of India's 22 official languages, piloted since April 2023 in rural Haryana.

10 of 22 official languages
Languages covered
171 of ~20,000 schemes
Government schemes covered
Jugalbandi — A Multilingual Generative-AI Chatbot Connecting Rural Indians to Government Welfare Schemes

Details

Maturity
Pilot
Promoter
AI4Bharat (IIT Madras) and OpenNyAI, with Microsoft Research India
Period
April 2023–present (pilot reach as most recently reported)
Keywords
generative AI, multilingual NLP, e-government, rural digital inclusion, government scheme access

Context

Jugalbandi is a generative-AI chatbot piloted since April 2023 in Biwan, in Haryana's Nuh district, designed to help rural Indians navigate roughly 20,000 government welfare schemes across India's central and state governments, most of which are inadequately publicised and documented only in English or Hindi. It was developed by AI4Bharat (a government-backed language-AI initiative at IIT Madras) and OpenNyAI (an open-source legal-tech collective), with reasoning models from Microsoft Research India via Azure OpenAI Service.

Activities

The system integrates Microsoft's GPT-4 models with Bhashini, India's national public speech-and-translation AI infrastructure, letting residents ask questions by voice or text in their native language and receive answers in that language even when the underlying scheme data exists only in English. Access is via WhatsApp and Telegram, chosen for their existing rural adoption, and the 'Jugalbandi Manager' platform is released as open-source software for adaptation to further languages or channels.

Results

As of the most recent public reporting, Jugalbandi covered 10 of India's 22 scheduled official languages and 171 of approximately 20,000 government schemes. No party has published usage volumes, resolution accuracy rates, or user-satisfaction metrics, and measured citizen impact on actual benefits access remains undocumented.

Conclusions

The evidence base rests on demonstrated technical reach rather than proven outcomes. Open-source release of the Jugalbandi Manager and reliance on India's public Bhashini infrastructure, rather than a proprietary closed system, are notable transparency strengths despite the absence of impact data.

Implementation

Indicative cost
Low (< €50k) — No budget figures disclosed; built on existing public Bhashini infrastructure and free messaging channels (WhatsApp, Telegram), suggesting relatively low direct cost.
Time to results
Medium (1–3 years) — Piloted from a single village (Biwan, Nuh district) starting April 2023, with coverage expanding gradually through the most recent (2026) reporting.
Staffing & skills
AI4Bharat (IIT Madras) language-AI researchers, OpenNyAI open-source legal-tech collective, Microsoft Research India engineers (Azure OpenAI Service)

Conditions for success

  • Reliance on existing public Bhashini speech/translation infrastructure rather than building translation capability from scratch
  • Use of high-adoption rural channels (WhatsApp, Telegram)
  • Open-source release of the Jugalbandi Manager enabling adaptation to new languages and channels

Common failure modes

  • No published usage volumes, resolution accuracy, or user-satisfaction metrics
  • Measured citizen impact on actual benefits access remains undocumented
  • Grassroots academic-NGO-corporate collaboration lacks a formally mandated government programme with disclosed oversight

Where it fits

Governance type
academic-NGO-corporate partnership operating alongside central/state welfare schemes
Scale
sub-national (single-village pilot expanding gradually)
Income level
lower-middle income

Commonly funded by

Philanthropic / foundation funding

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful