I think this will be extremely challenging to find good data on, and thus this will need to resolve based on bad data instead.

If Indian government provides some sort of "retraining fund" for job losses due to AI, then lots of losses will be attributed to AI. If public sentiment is against 'AI stealing jobs' then companies will attribute layoffs to other factors.

A large % of displacement won't happen through mass layoffs, it will be that companies just stop making openings for that role. It's extremely hard to quantify that sort of thing at a national level.

There will be plenty of free info available on this, but they will be opinion pieces that only pretend to be data-driven. That is, this is a topic that people want to read about, and where there's demand, someone will step up and write stories and make reports. If there's no easily available, good data for them to use, they will use bad data or no data. Who's gonna call them out as wrong, after all? Their made up or poor data will get cited by the next wave.

Even paywalled reports won't necessarily be accurate, just refined a bit better with more hewing to consensus -- whatever that ends up being based on.

And a poll? Just another step away from anything factual.