Literacy. Policy. Tutor. Measurement. In that order.

6 months on-site, 2 trainers, whole school. We don't just train teachers. We train students, teachers, staff, and administrators — over 6 months in your building. One shared vocabulary across the school, from the principal's office to the Class 8 classroom.

Why we lead with literacy and not the tutor: trust before product. Every other vendor sells you software. We sell you the capability to handle AI in your school — and the tutor only enters once that capability exists.

12-month engagement · two trainers on-site

Co-authored. Tool-agnostic. Signed by your leadership. Yours to keep. Deshika ships a 7-template AI Policy Starter Pack. Your staff customises each template in a 2-day workshop. Your leadership signs the final policy. The Deshika name never appears in the document.

The 7 templates: Acceptable Use Policy (students) · Teacher AI Use Policy · Student-Facing AI Tool Policy (tool-agnostic) · Parent Communication on AI · Data Privacy + Child-Safety Statement · AI in Assessment Policy · AI Incident Response Procedure.

The conflict-of-interest guard: our templates name categories of tools, never specific vendor names. The school's tool-authorisation decision happens separately — we'd be the tool that your policy authorises, but we'd never write a policy that names ourselves as the only approved tool. After 25 schools the templates open-source under Creative Commons.

The one AI tutor your policy can authorise. Child-safe by design, vernacular-native, DPDP-architected. Classes 8-10. The tutor enters the school only after literacy is done and the policy is signed.

Three things make this tutor different from free consumer AI: (a) Child-safety by design — a 12-layer architecture with India-tuned Bengali, Hindi, Assamese safety classifiers and a documented T1/T2/T3 distress-handoff protocol. (b) Vernacular by default — the tutor speaks the school's language, not just English-translated-to-Hindi. (c) DPDP-architected — India-hosted, Schedule-IV-aware, school as Data Fiduciary.

Per-student + per-teacher pricing · scoped at the first call

Outcome is proof, not promise. Third-party-audited learning-gain measurement (J-PAL or equivalent). We publish the methodology. The schools opt in. The data is anonymised. The report is public.

What we will not do: make board-result guarantees. India's CCPA Coaching Guidelines (2024) ban "guarantee X%" claims in education advertising, with ₹1.39 crore in penalties levied across 31 institutes since 13 November 2024. We'd refuse to make those claims regardless of regulation — they're how the post-BYJU's market got burned.

What we will do: publish the first credible measurement of co-authored-policy + literacy-program + sanctioned-tutor as a combined intervention in Indian K-12. Because nobody has run that experiment yet.