The record is in Marathi, handwritten, and stamped across the middle.

A seizure notice arrives from a village revenue officer in Thane district. Handwritten Marathi on a pre-printed form. Survey numbers in a hand that takes a while to read. A received stamp across one corner, sitting over the text. Photocopied at some point in its life, then scanned, then photographed on a phone because somebody needed it that afternoon.

This is not an unusual document. It is exactly the kind of document Indian lawyers deal with every day.

It is also a useful reminder that there is a part of legal AI that gets much less attention than reasoning or agents: seeing the material in the first place.

We have become accustomed to asking whether an AI system can understand a document, compare it with another document, or reason across a matter. But all of that assumes the document has first been turned into something the system can actually understand.

For Indian legal work, that is often the harder problem.

The record might be handwritten. It might be in Marathi, Tamil or Bengali. It might mix English with a regional language. A stamp might sit directly over the sentence that matters. A photocopy might be faint, skewed or partially cut off. A handwritten entry might only make sense when read against the printed field next to it.

These are not edge cases. They are the record. And this is where the difference between a general-purpose AI system and one built around Indian legal work starts to matter.

We have been building the ingestion layer at Jurisphere with these documents in mind. But the more important point is what happens after the document is read.

A translated land record should not become the end of a process. It should become part of the matter: something that can be read alongside the other records. That is increasingly where we think the real work in legal AI lies.

The models will continue to get better at reasoning. But better reasoning only matters if the system has access to the right material, can understand it in the form in which it exists, and can preserve enough of its context to know why it matters.