evidoria

← Back to browse

Good practice

Shiksha Copilot — Microsoft Research's AI Lesson-Planning Assistant for Government Schools in Karnataka, India

India · Bengaluru · See the India profile

Microsoft Research India and Sikshana Foundation built an AI copilot that drafts curriculum-aligned lesson plans in Kannada/English for Karnataka government-school teachers in under two minutes; a peer-reviewed 2026 study tracked its rollout to 1,043 teachers in 757 schools.

60-90 minutes to 60-90 seconds
Lesson plan drafting time (pilot phase, self-reported by teachers)
40 minutes to about 10 minutes
One teacher's lesson-planning time
1,043 teachers
Teachers reached (2026 ACM study, 757 schools across 35 districts)
40-70 students
Typical class size (Bengaluru pilot schools)

Details

Maturity
Scaling
Promoter
Microsoft Research India / Sikshana Foundation
Period
2023–ongoing (co-design pilot from Oct 2023; distribution phase 2024–2025)
Keywords
AI copilot, lesson planning, teacher support, generative AI, public education

Context

Shiksha Copilot is an AI-powered lesson-planning assistant built by Microsoft Research India in partnership with the Sikshana Foundation, a Bengaluru-based NGO focused on public education in Karnataka. It combines OCR, computer vision and generative-AI models to help government-school teachers assemble curriculum-aligned lesson plans - combining textbook content, videos, classroom activities and assessment tools - in Kannada and English, accessible via WhatsApp, Telegram or a web app.

Activities

The project began as a co-design pilot in October 2023 across more than ten government schools in and around Bengaluru, where teachers typically manage classes of 40 to 70 students. Sikshana Foundation, working with Karnataka government officials, subsequently recruited teachers from 757 public schools across 35 educational districts.

Results

Pilot teachers reported building a complete lesson plan in 60-90 seconds where it previously took 60-90 minutes; one teacher who used to spend up to 40 minutes on a single lesson plan cut that to about 10 minutes. A peer-reviewed 2026 study in the Proceedings of the ACM on Human-Computer Interaction documented the tool's distribution to 1,043 field teachers in Kannada- and English-medium schools, finding that it eased bureaucratic documentation workload, lowered teachers' reported stress and supported a shift toward more activity-based pedagogy.

Conclusions

The same study is candid that systemic pressures such as staffing shortages and administrative demands constrained how much the tool could shift broader pedagogical practice, and that the published evidence measures teacher time and workload rather than independently verified student learning outcomes.

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

Data sources

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

Similar practices you may find useful