Prompt Management
Autonify's Prompt Management feature allows administrators to customize and fine-tune the AI prompts used by various agents throughout the platform. This enables organizations to tailor AI behavior to match their specific standards, terminology, and requirements.
📹 Prompt Management Tutorial
Learn how to customize AI agent prompts to match your organization's standards and requirements.
Overview
Prompt Management allows administrators to customize the AI prompts used by various agents throughout Autonify. This enables you to tailor how AI agents interpret and analyze your data according to your organization's specific needs and terminology.
Accessing Prompt Management
- Navigate to the Settings panel in the Autonify platform
- Click on the Prompts tab (only visible to Administrators and Owners)
- You'll see a comprehensive table listing all available AI prompts
The Prompt Management interface displays all customizable AI prompts in a table format, showing the associated agent, prompt name, category, and last update date for each prompt
Available Prompts
The prompts panel displays all configurable AI prompts organized by:
- Agent Name - Which AI agent uses this prompt
- Prompt Name - What the prompt does
- Category - The type of analysis (analysis, validation, generation)
- Variables - Dynamic values used in the prompt
Common prompts you can customize include:
- Data Sensitivity Detection - Define what counts as sensitive data in your organization
- BI Audit Analysis - Customize how business intelligence investigations are performed
- Documentation Generation - Control how automatic documentation is created
- Semantic Grouping - Adjust how related data is organized and grouped
Editing Prompts
To customize a prompt:
- Click on any prompt in the list to open the editor
- The editor displays:
- Prompt Name - The name of the prompt you're editing
- Description - What this prompt is used for
- Prompt Content - The editable AI instructions
The Prompt Editor provides a full-featured text editing interface with syntax highlighting, line numbers, and automatic variable detection for customizing AI agent behavior
Editor Features
- Syntax Highlighting - Color-coded text for easier reading
- Line Numbers - Quick reference for specific sections
- Auto-detection - Variables in curly braces
{like_this}
are automatically recognized
Example: Customizing Data Sensitivity Detection
The Data Sensitivity Detection prompt is commonly customized to match organizational standards. You can define:
Your Sensitivity Levels
- High - What counts as highly sensitive in your organization
- Medium - Your moderately sensitive data types
- Low - Data that needs basic protection
Your Data Categories
Configure the prompt to recognize your specific data types:
- Customer information
- Financial records
- Employee data
- Business confidential information
- Public information
Customization Best Practices
1. Align with Organizational Standards
Modify prompts to match your organization's:
- Data classification standards
- Compliance requirements
- Industry-specific terminology
- Internal policies and procedures
2. Be Specific
When customizing prompts:
- Use clear examples of what you want
- Include your specific terminology
- Define boundaries clearly
3. Test Before Full Deployment
- Try your changes on a small dataset first
- Check the results match your expectations
- Adjust and refine as needed
Use Cases
Financial Services
Customize prompts to identify:
- Credit card numbers
- Account numbers
- Transaction data
- Risk indicators
Healthcare
Configure prompts to recognize:
- Patient identifiers
- Medical record numbers
- Diagnosis codes
- HIPAA-regulated data
Retail & E-commerce
Tailor prompts to detect:
- Customer purchase history
- Loyalty program data
- Inventory information
- Pricing strategies
Advanced Configuration
Using Variables
You can use variables in your prompts that will be replaced with actual values when the agent runs:
{data_source_name}
- Will be replaced with your database name{table_name}
- Will be replaced with the specific table being analyzed- Add any variable using curly braces and it will be automatically recognized
Adding Rules
You can include specific rules in your prompts:
- "If the data contains email addresses, mark it as highly sensitive"
- "If it's aggregated data with no individual records, mark as low sensitivity"
- "Financial account numbers should always be classified as high"
Access Requirements
Only Administrators and Owners can access and edit prompts. The Prompts tab will appear in Settings when you have the appropriate permissions.
What Happens When You Change a Prompt
- New agent runs will use your updated prompt
- The AI will follow your new instructions
- Results will reflect your customizations
- Previous results remain unchanged
Saving Your Changes
- Click the Save Changes button after editing
- You'll see a confirmation message when saved successfully
- Changes apply immediately to new agent runs
Important:
- Previous results won't be updated automatically
- Re-run agents on important data after making significant changes
Troubleshooting
If agents aren't producing expected results after modifying prompts:
- Check for clear, specific instructions in your prompt
- Ensure variables are properly formatted with curly braces:
{variable}
- Test with a small dataset first
- Try reverting to the original prompt if issues persist
Related Features
- Model Selection - Choose which AI models to use
- Data Sensitivity Classification - How sensitivity detection works
- Agent Configuration - Configure agent behavior