The staff room was sceptical, which is the honest place to start. Of the 34 teachers, four had used any AI tool before. The head of science summarised the mood: “We have seen a lot of technology arrive and then sit in a cupboard.”
Weeks one to three: friction
Adoption was slow and uneven. The teachers who moved first were not the youngest — they were the ones with the heaviest marking loads. Two of them started with worksheet generation and nothing else.
The turn
The shift came when one teacher shared a differentiated worksheet set in a staff meeting. It was not the technology that persuaded the room; it was seeing a colleague’s Sunday returned to them.
“I stopped thinking of it as software. It is closer to having an assistant who does the typing.” — Class 9 mathematics teacher
What did not work
We are not going to pretend the term was frictionless. Two things clearly fell short:
- Onboarding was too long. The initial training ran three hours. Teachers wanted twenty minutes and a reference card.
- Regional language output needed editing. Marathi explanations were serviceable but not idiomatic. Teachers corrected them, and those corrections have fed back into the model.
Where it stands
By October, 29 of 34 teachers were using at least one tool weekly. The most-cited benefit was not exam results — too early for that — but preparation time.