AI in EdTech is only as good as whether a student in a tier-2 or tier-3 town can actually use it. A demo that impresses in a metro classroom means nothing if it needs a fast connection, an expensive subscription, or fluent English to be useful.
Cost and language come first
Any AI feature filed under the Neo Bharat registry has to clear two bars before anything else: it has to be cheap enough that price is never the reason someone is left out, and it has to work in the language a student actually thinks in — Hinglish included, not just English.
Not overclaiming what the model can do
The bigger risk with AI in education is not that it is too weak — it is that it is oversold. A tool that confidently gives a wrong answer does more damage than one that says "I'm not sure." Every AI feature in the registry is built to be honest about its limits before it is built to look impressive.