> ## Documentation Index
> Fetch the complete documentation index at: https://docs.insight.nobly.dk/llms.txt
> Use this file to discover all available pages before exploring further.

# Smart search

> Find the right document fast in Nobly Insight — structured keyword search, cascading datasets, and AI-powered natural-language search and chat.

## Overview

Find the right document — fast — whether you know exactly what you're looking for or only have a vague idea. Nobly Insight combines traditional structured search with an AI-powered semantic layer, all governed by the same per-user permissions.

## Structured search

* Search across **hundreds of millions of documents** by keyword, date, document type, and combinations of metadata
* **Cascading datasets**: dependent dropdowns (e.g. customer → contract → policy) for guided searches and entry forms
* **Auto-fill keyword sets**: type a single value (such as a customer number) and the rest of the related metadata is filled in for you
* **External datasets**: pull autofill data from external systems (CRM, ERP, master data) without copying it
* **Saved searches** and personal/shared search folders

## AI-powered search

| Capability                      | What it does                                                                                                                         |
| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------ |
| **AI document search**          | Ask in natural language ("show me the latest credit decisions for this customer") and the platform finds the most relevant documents |
| **AI-generated search queries** | Convert a free-text question into a structured query you can refine and save                                                         |
| **AI Chat**                     | A conversational interface that lets users ask follow-up questions and dig deeper without re-typing context                          |
| **AI Summary**                  | Generate a concise summary of any document, or of a set of documents, in seconds                                                     |
| **AI Data Extraction**          | Extract structured data (e.g. fields from invoices, IDs, or forms) into searchable metadata                                          |

All AI search results respect each user's existing security — users only ever see results for documents they are permitted to access. AI processing runs on Nobly-owned infrastructure in the EU; see [How we use AI](/ai/introduction) for the full data and privacy model.

## How retrieval works

AI Search combines BM25 keyword ranking with vector-based semantic ranking, optionally enhanced by an LLM that rewrites the query before search and reranks results afterwards.

For ranking details, scoring, advanced settings, and tips, see:

<Card title="AI Search introduction" icon="magnifying-glass" href="/ai-search/introduction" horizontal>
  How results are ranked and scored, what each toggle does, and how to get the best results.
</Card>

<Card title="LLM reranker and query rewrite" icon="sparkles" href="/ai-search/llm-reranker-and-rewrite" horizontal>
  When the LLM rewrite and rerank stages run, what they output, and when to turn them off.
</Card>

## Where to read next

<Card title="Document management" icon="folder-open" href="/features/document-management" horizontal>
  How documents are captured, stored, edited, and governed before they reach search.
</Card>

<Card title="How we use AI" icon="shield-check" href="/ai/introduction" horizontal>
  The principles behind every AI feature: open-weight models, Nobly-owned infrastructure, and no third-party AI subprocessors.
</Card>