Prime Highlights:
- Anthropic formed its first AI lab partnership in India with Pratham to roll out an AI-powered formative assessment system across schools and women’s education programs.
- The initiative uses Anthropic’s Claude model to automate question generation, grading, and personalized feedback, addressing teacher workload challenges in large classrooms.
Key Facts:
- The tool is currently being tested with 1,500 students in 20 schoolsand is being adapted for over 5,000 women in Pratham’s Second Chance program, with full AI transition planned by 2026.
- Grading accuracy improved from around 30% to nearly 80%, while question generation now achieves about 90% alignment with Bloom’s Taxonomy standards.
Background:
Anthropic has partnered with Pratham Education Foundation to deploy an AI-based formative assessment tool across Indian schools. This is Anthropic’s first AI lab partnership with one of India’s biggest education nonprofits.
The initiative focuses on Pratham’s Anytime Testing Machine (ATM), powered by Anthropic’s Claude model. The system creates curriculum-aligned questions, grades handwritten responses and delivers personalized feedback. Students write answers on paper, upload images, and the system converts them into text before evaluation.
The tool is currently being tested with 1,500 students in 20 schools. Pratham is also adapting it for more than 5,000 women enrolled in its Second Chance program, which prepares learners for India’s grade 10 board exam. The broader program serves around 15,000 women and plans to fully shift to the AI-powered system by 2026.
Pratham developed the tool to solve a long-standing grading challenge. Many teachers manage classrooms of 60 or more students, limiting detailed feedback. The issue is more severe in programs serving women who left formal education and lack access to trained instructors.
Anthropic and Pratham teams worked together for several months to refine the system. Grading accuracy rose from around 30% to nearly 80% when compared to expert benchmarks. Question generation now achieves about 90% alignment with Bloom’s Taxonomy standards.
The platform gives feedback in both Hindi and English. Teachers play a key role by checking the AI-generated feedback before giving it to students.
Anthropic recently opened an office in Bengaluru to grow its presence in India. India is now the second-biggest market for Claude.ai, with high demand from developers and teachers. The two organizations plan to run more trials and may expand into other Global South regions.
