I run AI workshops for colleges. One of my favourites was a two-day, hands-on agentic AI workshop for first-year students, where they built a multi-agent business simulation in Google Antigravity. First-years, building agents, in two days. The college told us the feedback was very positive, and honestly, the energy in the room said the same. Here is what works.
5 ways teachers can use AI this week
- Lesson plans: give the topic, class level, duration and one activity you like.
- Practice questions at three difficulty levels, with answers.
- Explanations in simpler English or Tamil, Hindi and other languages. Always check the translation.
- Feedback drafts on student work, which you then edit.
- Admin: circulars, parent messages, timetables.
My one piece of advice for teachers: start with the task you find most boring, not the most impressive one. Question banks and circulars are where you will save time first.
Free or low-cost tools for educators
- Students: Google is offering Indian college students aged 18 and above a free 12-month Google AI Plus plan, verified through SheerID at one.google.com/ai-student. The offer must be redeemed by 31 December 2026. GitHub Copilot Student is free for verified students, including in India.
- Teachers: GitHub offers Copilot Pro free to verified teachers. Anthropic's Claude for Teachers is currently for US K-12 educators only, so it is not available to teachers in India yet.
- No budget at all: free ChatGPT, Claude and Gemini, or a local model with Ollama.
What about AI cheating and detectors?
AI detectors are less reliable than people assume. In a July 2026 test by Epoch AI, when AI imitated a specific author's style, leading detectors missed between 10% and 18% of the AI-written passages. And the same tools can wrongly flag honest students.
Better options:
- Ask for drafts, notes and a short viva.
- Grade the process and the reflection, not just the final text.
- Allow AI openly, and ask students to say how they used it.
For colleges: run hands-on workshops, not lectures
What I've seen work with students:
- Start with a real problem from their lives (a timetable, a club website).
- Build it with AI in the first session.
- Put it online, so they can share it.
- End with what can go wrong: accuracy, privacy, keys.
Lab setup questions (which tools, laptops, accounts) came up in about five of my workshop planning calls. Sort out accounts and installs before day one. A workshop that loses its first hour to logins never recovers its energy.