Copilot - Meetzy Docs

What is Copilot?

Copilot is an advanced AI-powered prompt editor integrated into the Advanced Editor. Instead of just giving you suggestions, Copilot directly edits and improves your agent’s system prompt based on your instructions using a sophisticated structured editing system. It analyzes your current prompt, generates targeted improvements, and applies precise changes automatically through an advanced edit plan protocol.

Copilot doesn’t just suggest - it edits with precision. The new structured editing system uses targeted operations like section replacement, content insertion, and smart modifications to transform your system prompt while maintaining its structure and quality.

Accessing Copilot

Copilot is located in the right panel of the Advanced Editor. Click the “Copilot” tab or press ⌘J to access it quickly.

Structured Editing System

Copilot now operates through an advanced EditPlan system that provides precise, structured modifications to your agent’s configuration:

Edit Operations
Diff Integration

Replace Section

replace_section: Completely replaces a targeted section with improved contentUses intelligent boundary detection to preserve document structure while updating specific portions.

Insert Content

insert_after: Adds new content after specified sections or markersPerfect for expanding instructions or adding new behavioral guidelines without disrupting existing content.

Append Section

append_section: Adds entirely new sections to the end of your promptIdeal for introducing new capabilities, guardrails, or specialized instructions.

Delete Section

delete_section: Removes outdated or redundant sections cleanlyMaintains document integrity while eliminating unnecessary content.

Simplified SSE Protocol (5 Events)

The new Copilot communication protocol uses a streamlined 5-event system for real-time feedback:

Event Type	Description	Visual Indicator
thinking	AI is analyzing your request	⏱️ Thinking timer with seconds counter
text	Streaming explanation content	📝 Real-time text appearing
edit	Structured edit operation	✏️ Diff preview with operation details
complete	Operation finished successfully	✅ Summary with metrics and next steps
error	Something went wrong	❌ Error message with suggested fixes

Extended Event Types for Parameters

Additional events handle configuration changes:

input_change: Modifications to input parameters
output_change: Updates to output settings
enrichment_set: Changes to output enrichment
evaluation_change: Evaluation criteria updates

Welcome Screen

When you first open Copilot, you’ll see a clean welcome interface designed for quick access to powerful editing capabilities:

Voice Input
Call Context Analysis

Enhanced Voice Commands: Use voice to naturally describe improvements

Start Recording

Hold Microphone Button: Press and hold the mic button in the input fieldDouble-Shift Shortcut: Press Shift twice quickly to start voice recording

Speak Naturally

Describe changes in conversational language:“Add greeting instructions for appointment scheduling calls”“The agent needs to better handle pricing objections based on call ABC-123”

Review Transcription

Voice input is converted to text - review and edit before sending

Release the mic button or press Shift again to stop recording. Voice works best in quiet environments with clear speech.

Smart Call Integration: Reference specific calls to inform improvementsUse the @call:ID syntax to provide concrete examples:

@call:abc123 The customer was confused about pricing. 
Improve the pricing explanation section to be clearer.

Copilot automatically loads call transcripts and uses conversation context to make targeted improvements.

How Structured Editing Works

The new Copilot operates through a sophisticated three-phase process with integrated diff management:

🔍 Analysis Phase

Deep Context Understanding: Copilot analyzes your system prompt structure, identifies improvement opportunities, and loads any referenced call contexts.

Parses document structure and sections
Loads @call:ID transcripts for context
Identifies patterns and improvement areas

⚙️ Edit Planning

Structured Operation Generation: Creates a precise EditPlan with targeted operations instead of wholesale rewrites.

Generates specific edit operations (replace, insert, append, delete)
Preserves document structure and formatting
Plans minimal, focused changes

✏️ Diff Review & Application

Staged Changes with Review: All modifications appear in diff view for review before application.

Visual diff highlighting shows exactly what changes
Accept/reject controls for each operation
Maintains version history for rollbacks

Parameter & Configuration Changes

Copilot now intelligently suggests changes to input parameters, output settings, and evaluations that appear directly in their respective configuration panels:

Input Parameter Operations
Output Configuration
Evaluation Criteria

Smart Parameter Suggestions based on conversation analysis and agent needs:

Contextual Generation

Copilot analyzes your agent’s role and referenced calls to suggest relevant input parameters:

Role-based: Appointment schedulers get calendar preferences
Call-driven: Missing information from transcripts becomes parameters
Industry-standard: Common fields for your agent’s domain

Change Operations

Three types of parameter changes:

🟢 Add: New parameters with proper validation
🟡 Modify: Updates to existing parameter configs
🔴 Remove: Clean up outdated or redundant parameters

Visual Review System

Purple highlighting in Input Parameters panel shows pending changes:

Operation badges (Add/Modify/Remove)
Parameter details and descriptions
Individual accept/reject buttons
Bulk action controls

Intelligent Output Field Suggestions for comprehensive data capture:

@call:abc123 The agent collected the customer's name and interest level 
but we're missing their preferred contact method. Add output fields to capture this.

Copilot suggests contextually relevant output parameters:

Contact preferences (email, phone, text)
Lead qualification scores
Next step requirements
Custom fields based on conversation patterns

Visual Change Management System

Purple Highlight Protocol

When Copilot generates suggestions for configuration changes, distinctive visual indicators guide your review process:

Pending Changes Banner
Change Cards
Cross-Panel Integration

Purple-highlighted banners appear at the top of relevant panels:

Copilot sparkle icon with “Copilot Suggestions” title
Change summary: “3 new parameters, 1 evaluation”
Bulk controls: Accept All/Reject All buttons
Panel navigation: Direct links to affected configuration areas

Change Type Indicators

Change Type	Badge Color	Operation	Description
Add	🟢 Green	ADD	New parameters, evaluations, or settings
Modify	🟡 Yellow	MODIFY	Changes to existing configurations
Remove	🔴 Red	REMOVE	Deletion of outdated or redundant items

Accepting and Rejecting Suggestions

Granular Change Control

Individual Review

Examine each suggestion carefully:

Review operation type and parameter details
Check impact on your agent’s behavior
Verify alignment with your specific use case

Selective Application

Accept only relevant changes:

Click “Accept” on specific suggestions you want to keep
Click “Reject” to dismiss unsuitable suggestions
Changes apply immediately to your configuration

Live Configuration Updates

See results instantly:

Accepted changes appear in your panels immediately
Rejected changes are removed from pending list
Configuration panels update in real-time

Bulk Operations

Efficient management of multiple changes:

Accept All - Apply all pending suggestions simultaneously
Reject All - Dismiss all pending suggestions at once

Bulk actions are permanent and cannot be undone. Review individual suggestions first to ensure they align with your agent’s requirements before using “Accept All”.

Change Persistence & History

Session persistence: Pending changes survive editor sessions until resolved
Auto-save decisions: Your accept/reject choices are automatically preserved
Version tracking: All applied changes are logged for potential rollback
Change history: Full audit trail of Copilot modifications

Voice Recording & Analysis

Enhanced voice input capabilities for natural prompt editing:

Voice Command System
Call Context Integration

Activation Methods

Two ways to start voice recording:

Hold microphone button in the input field
Double-press Shift to begin hands-free recording

Natural Speech Processing

Speak conversationally:

Describe changes as you would to a colleague
Reference specific problems or improvements needed
Use call references naturally: “In call A-B-C-1-2-3, the customer was confused…”

Transcription & Review

Quality control before processing:

Voice converts to text automatically
Review and edit transcription if needed
Send refined request to Copilot

Powerful call reference system for context-driven improvements:

Using Call References

@call:abc123 The agent failed to handle the budget question. 
Add instructions to better address pricing concerns.

When you reference a call:

Copilot automatically loads the full transcript
Analyzes conversation flow and failure points
Uses specific context to generate targeted improvements

Call Context Indicators

Referenced calls appear with rich metadata:

Clickable call references: @call:abc123 format
Hover details: Duration, outcome, timestamp
Transcript access: Click to view full conversation

Voice Input Best Practices

Recording Environment
Effective Voice Commands
Voice vs Text Guidelines

Optimize for quality transcription:

Use quiet spaces with minimal background noise
Position close to your device’s microphone
Speak at normal, conversational pace
Enunciate clearly without over-articulating

Examples of natural voice instructions:

"Add greeting instructions for appointment scheduling calls"

"Reference call A-B-C-1-2-3. The customer was confused about pricing. 
Improve the pricing explanation section to be much clearer."

"Make the entire prompt sound more friendly and conversational 
while keeping it professional"

Use Voice When	Use Text When
Making quick, natural improvements	Writing complex technical instructions
Referencing call transcripts	Including specific code or formatting
Describing tone or style changes	Making precise edits to existing content
Brainstorming improvements	Adding URLs or technical specifications
Explaining context from calls	Copying exact text or templates

Call Context Analysis

Reference specific calls to inform targeted prompt improvements:

Identify Problem Calls

Find calls where the agent underperformed:

Review call transcripts in your Calls dashboard
Look for conversations with poor outcomes
Note specific failure points or confusion

Reference in Copilot

Use @call:ID syntax to provide context:

@call:xyz789 Improve how the agent handles "I need to think about it" responses

Copy call ID from dashboard or URL
Describe specific improvement needed
Let Copilot analyze the full conversation context

Context-Driven Improvements

Copilot uses call data to make targeted changes:

Analyzes conversation flow and failure points
Identifies missing instructions or unclear guidance
Generates specific improvements based on actual problems
Creates contextual parameter suggestions if data is missing

Call Context Features

Automatic transcript loading: Call content loads seamlessly
Metadata integration: Duration, outcome, participant info
Pattern recognition: Identifies common issues across referenced calls
Smart parameter generation: Suggests missing input/output fields based on call content

Advanced Editing Capabilities

Copilot can modify any aspect of your agent configuration with surgical precision:

Prompt Structure
Configuration Changes

Document-level improvements:

Element	Capabilities
Instructions	Add new behaviors, modify existing rules
Tone & Style	Adjust formality, friendliness, professionalism
Organization	Reorganize sections, add headers, improve flow
Guardrails	Add safety limits, scope boundaries
Scenarios	Add handling for specific situations
Language	Improve clarity, remove ambiguity

Next Step Suggestions

After completing changes, Copilot provides intelligent follow-up recommendations:

Using Suggestions Effectively

One-click application: Click suggestions to auto-populate input field
Customizable requests: Modify suggestions before sending
Combination strategies: Merge multiple suggestions into comprehensive improvements
Strategic dismissal: Remove suggestions that don’t fit your use case

Conversation History & Context

Build sophisticated improvements through multi-turn conversations:

User: Add instructions for handling price objections

Copilot: [Makes targeted changes with diff preview]

User: Good, but make it less aggressive and more consultative

Copilot: [Refines the approach with updated diff]

User: Perfect, now add a fallback for persistent objections

Copilot: [Adds additional escalation handling]

Copilot maintains conversation context across multiple requests, allowing for iterative refinement. Start a new conversation (using the menu option) when switching to completely different topics.

Common Editing Tasks

Enhanced Behavioral Instructions

Add instructions for the agent to always confirm 
the customer's phone number before ending the call, 
and provide a backup email if the phone number is invalid

Context-Driven Scenario Improvements

@call:abc123 Improve how the agent handles when 
customers ask to speak with a manager - the current 
approach seemed to frustrate the customer

Comprehensive Style Adjustments

Make the entire prompt more conversational and less formal. 
The agent should sound like a helpful friend, not a robot, 
while maintaining professionalism for business contexts.

Integrated Parameter & Output Management

Add input parameters to capture customer budget range and timeline, 
plus output fields to track lead qualification scores and next steps

Performance-Driven Evaluations

Create evaluations to measure: conversation quality (1-10 scale), 
goal achievement (percentage), customer satisfaction (survey-based), 
and proper procedure following (checklist compliance)

Contextual Guardrail Development

@call:xyz789 The agent made promises about delivery dates. 
Add clear boundaries: never discuss competitor products 
or make specific delivery commitments without manager approval

Call-Driven Parameter Updates

@call:def456 The customer mentioned their budget repeatedly 
but we didn't capture it. Add input and output parameters 
for budget tracking with proper validation rules.

Keyboard Shortcuts

Shortcut	Action
⌘J	Open/focus Copilot panel
Shift + Shift	Start/stop voice recording
⌘Enter	Send message to Copilot
Escape	Close Copilot suggestions or cancel recording
⌘3	Navigate to Input Parameters panel
⌘6	Navigate to Evaluations panel
⌘7	Navigate to Output Settings panel
⌘G	Navigate to Git Changes panel

Best Practices

Be Specific About Changes

Use precise, actionable descriptions:Good: “Add a 3-step refund request handling process with clear decision criteria”Vague: “Improve customer service handling”Specific requests generate more targeted and useful improvements.

Leverage Real Call Context

Reference actual problematic calls for concrete improvements:

Use @call:ID syntax to provide specific failure examples
Let Copilot analyze real conversation data
Address actual problems rather than hypothetical scenarios

Review Parameter Suggestions Strategically

Don’t automatically accept all configuration changes:

Check Input Parameters and Output Settings panels for purple highlights
Evaluate each suggestion against your specific use case
Consider the impact on conversation flow and data collection

Use Voice for Natural Requests

Voice input excels for conversational improvement descriptions:

“The agent needs to collect customer budget and timeline preferences”
“Based on call ABC-123, improve how we handle pricing questions”
Let natural speech guide more intuitive improvements

Follow Visual Notification Cues

Purple indicators guide you to pending changes:

Check all panels with notification dots
Don’t miss suggestions in configuration areas
Use cross-panel navigation to review comprehensively

Test Comprehensively After Major Changes

Validate improvements before deploying:

Use Playground to test new behaviors
Run regression tests to ensure quality maintenance
Verify parameter changes work with real data

Iterate Incrementally

Make focused improvements rather than massive overhauls:

Address one problem area at a time
Test changes before adding more complexity
Build understanding of what works through smaller steps

Next Steps

Input Parameters

Configure input parameters with AI suggestions from Copilot

Output Settings

Set up output fields and webhooks with intelligent suggestions

Evaluations

Create performance measurements with Copilot recommendations

Playground

Test your improved prompts and parameter changes

Tests

Create automated tests to validate Copilot improvements

Version History

Track all changes from Copilot’s structured editing system

Getting Started

Agents

Advanced Editor

Voice Library

Calls

CRM

Account

​What is Copilot?

​Accessing Copilot

​Structured Editing System

​Simplified SSE Protocol (5 Events)

​Extended Event Types for Parameters

​Welcome Screen

​How Structured Editing Works

​Parameter & Configuration Changes

​Visual Change Management System

​Purple Highlight Protocol

​Change Type Indicators

​Accepting and Rejecting Suggestions

​Granular Change Control

​Bulk Operations

​Change Persistence & History

​Voice Recording & Analysis

​Using Call References

​Call Context Indicators

​Voice Input Best Practices

​Call Context Analysis

​Call Context Features

​Advanced Editing Capabilities

​Next Step Suggestions

​Using Suggestions Effectively

​Conversation History & Context

​Common Editing Tasks

​Enhanced Behavioral Instructions

​Context-Driven Scenario Improvements

​Comprehensive Style Adjustments

​Integrated Parameter & Output Management

​Performance-Driven Evaluations

​Contextual Guardrail Development

​Call-Driven Parameter Updates

​Keyboard Shortcuts

​Best Practices

​Next Steps

Input Parameters

Output Settings

Evaluations

Playground

Tests

Version History

What is Copilot?

Accessing Copilot

Structured Editing System

Simplified SSE Protocol (5 Events)

Extended Event Types for Parameters

Welcome Screen

How Structured Editing Works

Parameter & Configuration Changes

Visual Change Management System

Purple Highlight Protocol

Change Type Indicators

Accepting and Rejecting Suggestions

Granular Change Control

Bulk Operations

Change Persistence & History

Voice Recording & Analysis

Using Call References

Call Context Indicators

Voice Input Best Practices

Call Context Analysis

Call Context Features

Advanced Editing Capabilities

Next Step Suggestions

Using Suggestions Effectively

Conversation History & Context

Common Editing Tasks

Enhanced Behavioral Instructions

Context-Driven Scenario Improvements

Comprehensive Style Adjustments

Integrated Parameter & Output Management

Performance-Driven Evaluations

Contextual Guardrail Development

Call-Driven Parameter Updates

Keyboard Shortcuts

Best Practices

Next Steps