White Paper

Beyond Retrieval:
The Case for Emergent Document Intelligence

A technical exploration of why the next generation of document intelligence isn't about better search — it's about discovering the metadata that makes search intelligent in the first place.

Includes Real Findings From

510
commercial contracts
CUAD · B2B · 15+ industries
1,910
consumer agreements
CFPB · B2C · credit cards
0
configuration changes
Same engine. Different domains.

What's Inside

1
The Limits of the Current Paradigm
What RAG does well, what it cannot do, and why the analyst's problem is that the corpus knows more than any query.
2
A New Architecture for Document Intelligence
Three layers, three questions. Why each is insufficient alone — and how they compose into something qualitatively different.
3
The Retrieval Foundation
Semantic search and its limitations. Hierarchical summarization. Adaptive metadata that emerges from content type.
4
The Relationship Graph
Entity and relationship extraction at scale. What graph-based retrieval enables that vector search cannot. The temporal dimension.
5
The Emergent Metadata Engine
The central claim: unsupervised schema discovery. The four waves. Temporal intelligence. What 'projecting schema onto documents' means.
6
The New Class of Questions
Before Tessera vs. after Tessera. Real findings from 2,420 contracts across two domains. The trust problem: intelligence must be evidenced, not asserted.
7
What Tessera Does Not Do
Failure modes, confidence boundaries, and the dark hallway principle. Why admitting limits builds more trust than omitting them.
8
Conclusion: The Mosaic
Intelligence from documents isn't retrieved — it's assembled. Your data is full of tiles.

Get the paper when it's published.

We'll send it once. No spam. No sequence.

Intelligence from documents isn't retrieved — it's assembled.
Every record is a tile. See the whole picture.