> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://support.contracthero.com/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# How to write a prompt in ContractHero

ContractHero’s **Contract Analysis** feature leverages AI to automate and streamline the process of extracting key data from contracts. Here's an in-depth look at how it works and how to optimize the prompts for better results:

### Automated Data Extraction:

ContractHero’s AI-powered tool extracts critical contract information such as **dates**, **parties involved**, and **financial terms** from uploaded documents. This process saves time and eliminates manual data entry.

### Contract Analysis Process:

1. **Document Upload**: The user uploads the contract document into the system.

3. **OCR (Optical Character Recognition)**: The system first uses OCR technology to convert the text in the contract into a machine-readable format, ensuring that even scanned or image-based documents are accessible for analysis.

5. **Custom Field Data Extraction**: The AI extracts information based on predefined **custom fields**. These fields are tailored to the specific category of the contract, ensuring that only the relevant data is pulled for that particular contract type.

7. **AI Request Breakdown**:
    
* **Contract document text**: The raw text extracted from the contract.
    
* **Field name**: The designated label for the data being requested (e.g., "Start Date").
    
* **Field type**: The format of the expected data (e.g., date, number, text).
    
* **Field prompt**: A clear instruction that defines what the AI is looking for within the contract.

5. Based on this structure, the AI delivers a result by scanning the contract document and matching the information with the prompt and field type.

### Context and Setup:

This information is already provided by ContractHero and does not need to be included in the custom field prompt.

* **Task Guidance**: The AI model receives essential details about the task at hand, such as which contract category it is working with, ensuring the extraction is tailored to the specific document type.
* **Field Formatting**: Data extracted is formatted according to the defined **field type** (e.g., date formats, numerical values).
* **Organizational Context**: The AI is also aware of the organization’s name to accurately identify the relevant clauses and sections in the contract where it should focus its analysis.
* **Bilingual Support**: ContractHero provides bilingual support, making it capable of analyzing contracts in both **German and English**.

### Effective Prompt Writing:

To ensure that the AI returns accurate results, it's crucial to write clear and precise prompts. Below are tips for optimizing your prompts:

1. **Be Specific in Instructions**:
    
* **Bad example**: “What date is written in the contract?”
    
* **Good example**: “What start date is written in the contract?”
    
* Clear, specific instructions help the AI focus on the right information.

2. **Provide a Detailed Description of the Value**:
    
* **Bad example**: “When does the contract start?”
    
* **Good example**: “What is the start date of the contract? Often found under 'contract term' or 'contract duration'."
    
* The more context you provide, the better the AI can locate the right data.

3. **Ensure Consistency in List Values**:
    
* **Bad example**: The list of options contains "Car" and "Coffee machine", but the prompt asks: "Select cars for a vehicle leasing contract"
    
* **Good example**: ![](https://t6640275.p.clickup-attachments.com/t6640275/4cbeeddc-eb57-4a5d-85c0-52d8f6eabb58/image.png)
    
* The prompt and the list of acceptable values should match exactly to avoid confusion and ensure the AI extracts the correct response.

4. **Field Name Alignment**:
    
* **Bad example**: The field name is "payment" but the prompt asks about "salary".
    
* **Good example**: Ensure the field name aligns with the prompt. If the field is "payment", the prompt should refer to “payment” and not “salary”.
    
* Consistent terminology between the field name and prompt ensures accurate data extraction.

5. **Use of examples**
    
* **Example**: "What is the termination date of the contract? This is often listed at the end of the agreement or in Section 8 titled 'Termination.'"