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

# Document Intelligence

> How we understand your documents

# Document Intelligence

Every document is unique, but legal documents follow patterns. Our system
understands these patterns and uses that knowledge to provide smarter
validation, better risk detection, and more accurate editing.

## How We Understand Documents

When you upload a document, we don't just process text. We understand:

<CardGroup cols={2}>
  <Card title="Document Type" icon="file-contract">
    Is this an NDA, MSA, SOW, or something else?
  </Card>

  <Card title="Structure" icon="list-tree">
    What sections exist and how are they organized?
  </Card>

  <Card title="Key Parties" icon="users">
    Who are the parties to this agreement?
  </Card>

  <Card title="Critical Terms" icon="star">
    What are the key dates, terms, and provisions?
  </Card>
</CardGroup>

***

## Document Type Recognition

We automatically identify the type of document you're working with:

| Document Type  | We Look For                                                               |
| -------------- | ------------------------------------------------------------------------- |
| **NDA**        | Confidentiality definitions, permitted disclosures, return of information |
| **MSA**        | Services descriptions, payment terms, change order provisions             |
| **SOW**        | Project scope, deliverables, milestones, acceptance criteria              |
| **Employment** | Position, compensation, benefits, restrictive covenants                   |
| **SaaS**       | Access rights, SLA, data protection, service levels                       |
| **Purchase**   | Goods/services, pricing, delivery, acceptance                             |
| **Lease**      | Premises, rent, term, maintenance obligations                             |
| **Amendment**  | Recitals, specific modifications, ratification                            |

<Note>
  Document type affects how we validate, what risks we look for, and what
  sections we expect to find. An NDA has different standards than an MSA.
</Note>

***

## Expected vs. Actual Structure

### What We Expect

Each document type has expected sections. For example, an NDA typically includes:

<Check>Definitions</Check>
<Check>Confidential Information</Check>
<Check>Obligations</Check>
<Check>Exceptions/Exclusions</Check>
<Check>Term and Termination</Check>
<Check>Return of Information</Check>
<Check>Remedies</Check>
<Check>Miscellaneous Provisions</Check>

### What We Find

We analyze your actual document and identify:

* **Present sections** that match expected structure
* **Missing sections** that are typically included
* **Additional sections** specific to your document
* **Section boundaries** for targeted editing

### Why This Matters

When a standard section is missing, it might be:

* Intentional (not needed for this deal)
* An oversight (should be added)
* Different naming (called something else)

We note these findings without blocking your work—just giving you visibility.

***

## Key Information Extraction

### Parties

We identify the parties to the agreement:

```
Parties Detected:
- Acme Corporation (the "Company")
- John Smith (the "Consultant")
```

### Key Dates

We extract important dates:

| Date Type      | Value             |
| -------------- | ----------------- |
| Effective Date | January 1, 2024   |
| Term           | 24 months         |
| Expiration     | December 31, 2025 |

### Critical Terms

We identify key terms that affect the agreement:

```
Key Terms:
- Governing Law: Delaware
- Arbitration: Yes, AAA Rules
- Confidentiality Period: 5 years
- Non-Compete: 2 years, 50-mile radius
```

***

## How Classification Improves Editing

### Better Validation

When we know document type:

* We check for section-specific requirements
* We apply appropriate risk thresholds
* We verify expected provisions are present

### Smarter Risk Detection

Different documents have different risks:

* **NDA**: Focus on confidentiality scope, duration, permitted disclosures
* **MSA**: Focus on liability caps, indemnification, IP provisions
* **Employment**: Focus on compensation, non-compete, termination rights

### More Accurate Editing

Understanding document structure means:

* Section references are understood in context
* Related provisions are updated consistently
* Cross-references remain accurate

***

## Confidence Levels

Our classification includes a confidence score:

| Confidence    | Meaning                                         |
| ------------- | ----------------------------------------------- |
| **90%+**      | Very clear document type, strong pattern match  |
| **70-89%**    | Likely document type, most indicators present   |
| **50-69%**    | Possible document type, some indicators present |
| **Below 50%** | Unclear type, classified as "general"           |

<Info>
  When confidence is lower, we apply broader validation rules to ensure nothing
  is missed. We err on the side of more thorough review.
</Info>

***

## Automatic Tier Selection

Document classification determines validation tier:

| Classification                    | Validation Tier | Why                      |
| --------------------------------- | --------------- | ------------------------ |
| Legal contract (high confidence)  | Legal           | Full 5-pass validation   |
| Legal contract (lower confidence) | Legal           | Err on side of caution   |
| Non-legal document                | Standard        | Efficient processing     |
| Complex/high-value                | Enterprise      | Maximum rigor (optional) |

<Note>
  You can always override automatic tier selection if you need more or less
  validation for a specific document.
</Note>

***

## Supported Document Types

<AccordionGroup>
  <Accordion title="Non-Disclosure Agreements (NDAs)">
    Confidentiality agreements, mutual NDAs, unilateral NDAs, confidential
    disclosure agreements
  </Accordion>

  <Accordion title="Master Services Agreements (MSAs)">
    Professional services agreements, consulting agreements, master agreements
  </Accordion>

  <Accordion title="Statements of Work (SOWs)">
    Work orders, project agreements, service orders, task orders
  </Accordion>

  <Accordion title="Employment Agreements">
    Offer letters, employment contracts, executive employment agreements,
    contractor agreements
  </Accordion>

  <Accordion title="SaaS Agreements">
    Software license agreements, subscription agreements, cloud services
    agreements
  </Accordion>

  <Accordion title="Purchase Agreements">
    Sales agreements, procurement contracts, supply agreements, vendor
    agreements
  </Accordion>

  <Accordion title="Lease Agreements">
    Commercial leases, equipment leases, real property leases
  </Accordion>

  <Accordion title="Amendments">
    Contract amendments, modifications, change orders, supplements
  </Accordion>
</AccordionGroup>

***

<CardGroup cols={2}>
  <Card title="Risk Detection" icon="shield-halved" href="/legal-ai/risk-detection">
    How classification enables smart risk detection.
  </Card>

  <Card title="Multi-Pass Review" icon="list-check" href="/legal-ai/multi-pass-review">
    How we validate based on document type.
  </Card>
</CardGroup>