Autopilot Classic

You are here. Triggers after each call/email to analyze its content in real-time.

Retropilot

Triggers on a Salesforce event and analyzes all historical conversations.

Autopilot Batch

Triggers manually to run on a list of records all at once.

​
Setting Up Autopilot Classic

This guide provides step-by-step instructions for configuring Momentum’s classic Autopilot feature to automatically extract and update Salesforce data from your sales calls.

​
How Autopilot Works

Autopilot triggers after a single call or email, analyzes its content, and updates a Salesforce field.

​
Prerequisites

Before configuring Autopilot, ensure you have:

​
Step-by-Step Configuration

​
Step 1: Access the Autopilot Section

  1. Log in to your Momentum admin dashboard
  2. Navigate to the Autopilot section in the left sidebar
  3. Click “New Autopilot Extraction” to begin the setup process

​
Step 2: Select Salesforce Object and Field

Choose the Target Object:
  • Opportunity: Extract data to opportunity fields
  • Account: Extract data to account fields
  • Lead: Extract data to lead fields
  • Event: Extract data to event fields
  • Custom Objects: Extract to other object fields by navigating from one of the default objects above
Select the Target Field:
  • Standard Fields: Choose from standard Salesforce fields
  • Custom Fields: Select custom fields you’ve created
  • Picklist Fields: Automatically limit outputs for fields with predefined values
  • Text Fields: Automatically limit outputs for free-form text fields and their max length and formatting
  • Number Fields: Automatically limit outputs for numeric data
  • Boolean Fields: Automatically limit outputs for yes/no answers
Example Configuration: 
Object: Opportunity
Field: Pain_Points__c (Custom Text Field)

​
Step 3: Write Your AI Prompt

The AI prompt is the instruction that tells Momentum what data to extract from calls. Prompt Best Practices:
  • Be Specific: Clearly define what information you want extracted
  • Use Questions: Frame prompts as questions the AI should answer
  • Include Context: Reference the field name and expected format
  • Set Boundaries: Define what should and shouldn’t be included
Example Prompts: For Pain Points: 
"What are the main pain points or challenges the customer mentioned during this call? Focus on business problems they're trying to solve. Return a concise list of 2-3 key pain points."
For Budget Information: 
"What is the customer's budget range or budget constraints mentioned in this call? Include specific amounts, ranges, or budget-related discussions. If no budget information is discussed, return 'Not discussed'."
For Timeline: 
"What is the customer's timeline or urgency for implementing a solution? Include specific dates, quarters, or urgency indicators mentioned. If no timeline is discussed, return 'Not discussed'."
For Competitive Information: 
"Which competitors did the customer mention during this call? List specific competitor names and any context about their current solutions or evaluation status."

​
Step 4: Configure Salesforce Conditions

Set conditions to determine when this extraction should run. Example Conditions: 
Opportunity Stage: Discovery, Qualification
Account ARR: > $50,000

​
Step 5: Configure Save Behavior

Choose how the extracted data should be saved to Salesforce. Save Options:
  • Confirm to Write: Show extraction in Slack or web app for manual review before saving
  • Write if Empty: Only save if the field is currently empty, confirm to write otherwise
  • Automatic Write: Automatically save all extractions if new data is identified
Balancing Automation and Verification While manual verification may seem like the safer approach, our extensive field experience has shown that requiring rep engagement often becomes the biggest roadblock to maintaining high-quality, consistent data. Sales reps are less likely to consistently review and approve extractions, which leads to gaps in data collection and outdated information. Instead, we strongly recommend pushing towards “Write If Empty” for most extractions after your initial testing period. This approach strikes the optimal balance between maintaining data integrity and maximizing the benefits of automation. For best results, we suggest a simple progression: begin with “Confirm to Write” during the first 1-2 weeks while you validate the extraction quality. Once you’re confident in the results, transition to “Write If Empty” as your standard operating mode. For highly reliable extractions where the data changes frequently, you may eventually consider using “Automatic Write.” Remember, focusing your energy on crafting precise, well-defined prompts will typically yield better results than relying on manual verification processes.

​
Step 6: Test Your Extraction

  1. Preview Your Extraction: Click the “Preview” button before creating to test your prompt against past calls
  2. Review Sample Results: Examine how your prompt performs on historical call data without affecting Salesforce or Slack
  3. Refine the Prompt: Adjust the AI prompt based on preview results
  4. Create Live Extraction: Once satisfied with preview results, create the extraction for production use
Autopilot Configuration Interface

​
Advanced Configuration

​
Multiple Extractions for the Same Field

You can create multiple extractions for the same field with different conditions: Discovery Extraction:
  • Field: Pain_Points__c
  • Conditions: Opportunity Stage = “Discovery”
  • Prompt: Focus on initial pain point discovery
Qualification Extraction:
  • Field: Pain_Points__c
  • Conditions: Opportunity Stage = “Qualification”
  • Prompt: Focus on validated and prioritized pain points

​
Context-Aware Extractions

Configure extractions that adapt based on call context: Stage-Based Prompts:
  • Use different prompts for different opportunity stages
  • Adjust extraction sensitivity based on deal maturity
  • Include stage-specific context in prompts
Account-Based Rules:
  • Different extractions for enterprise vs. SMB accounts
  • Industry-specific prompts and conditions
  • ARR-based extraction rules

​
Best Practices

​
Prompt Engineering

  1. Start Simple: Begin with basic prompts and add complexity
  2. Test Iteratively: Refine prompts based on actual call data
  3. Use Examples: Include examples in prompts when possible
  4. Be Specific: Avoid vague language that could lead to inconsistent results
  5. Consider Context: Account for different call types and stages

​
Field Selection

  1. Prioritize High-Value Fields: Focus on fields that impact forecasting and reporting
  2. Consider Data Quality: Choose fields where manual entry is error-prone
  3. Start with Text Fields: Text fields are easier to configure than picklists
  4. Validate Picklist Mappings: Thoroughly test picklist field extractions

​
Troubleshooting

​
Common Issues

Extractions Not Running:
  • Check Salesforce conditions are correctly configured
  • Verify calls are being recorded and processed
  • Ensure AI license seats are properly assigned
Incorrect Data Extraction:
  • Review and refine AI prompts
  • Check that the right calls are being processed
  • Validate field mappings and conditions
Poor Data Quality:
  • Improve prompt specificity and clarity
  • Add examples to prompts for better guidance
  • Consider using “Confirm to Write” for manual review
  • Review extraction logs for patterns
  • Refine prompts based on error analysis
  • Consider breaking complex extractions into simpler ones

​
Support

If you encounter issues during configuration, contact our support team: