What happens to a document after you add it 

Every document goes through the same four stages. Typical total time is 30–90 seconds; long or scanned documents take longer.

The four stages

1. Classification: a model decides what kind of document this is and records how confident it was. Low confidence flows into review later.
2. Text extraction: the readable content is pulled out. Scanned PDFs go through image-based text recognition here.
3. Structured extraction: for leases, amendments and rent rolls only, the named facts are pulled into database fields: parties, dates, rents, areas, escalation schedules, unit rows.
4. Embedding and indexing: the content is indexed so it can be searched by meaning rather than by keyword. This is what lets Sunday find "the clause about early termination" without those exact words appearing.

Once all four are complete the document is fully usable and its extracted facts are available to validation and review.

What classification is, and is not

Three things are easy to conflate and worth separating for a customer:

• Classification labels a file. A correct label does not prove complete extraction.
• Extraction turns content into fields. A successful extraction does not prove a valid business result.
• Validation applies rules to what was extracted. A passed validation means the implemented rules found no issue in the evidence they could assess, not that the lease is fine.

Structured extraction is deep for contracts, amendments and rent rolls. Other document types are classified, text-extracted and searchable, but their facts are not pulled into fields.

Tips & best practices

• Status badges do not always refresh by themselves. A document stuck on "Classifying…" for more than a couple of minutes is usually finished; reload the page.
• Extracted facts do not become official portfolio data immediately. Anything uncertain goes to Reviews, and cross-document contradictions surface in Rent Roll Validation. That gate is deliberate; it is the reason the numbers can be trusted.