Skip to main content
Building an assistant in Envole is like creating a comprehensive job description and work environment for a digital employee. You’ll define their role, provide resources, and set up workflows just as you would for a new team member.

​
Getting Started with the Workflow Editor

To start building your assistant, find your way to the Workflows Editor:
  • Click the Create New Assistant button on the Home tab, or
  • Edit an existing workflow from the Assistants tab

​
Understanding the Interface

Once you’re in the editor, you’ll see:
  • Top navigation: Options to view Apps and Analytics
  • Toolbar (top right): Run, Save, and Edit workflow controls
  • Left sidebar: Components for building your workflow
  • Main canvas: Where you’ll build your assistant workflow
The left sidebar contains five types of components that work together to create your automated workflow:
  • Profile: The core identity and job description
  • Inputs: How information enters the workflow (e.g., customer questions)
  • Activities: How you want the assistant to process information (e.g., analyze sentiment)
  • Tools: External systems and integrations (e.g., email, CRM)
  • Outputs: The final results delivered to users (e.g., formatted responses)

​
Understanding How Components Work Together

Think of these components as different aspects of defining a complete work role: Profile is your assistant’s job description - it defines who they are, what they do, how they behave, and what they know. Inputs & Activities are the implementation details - Inputs determine how work comes to your assistant, while Activities define the specific tasks they perform. Tools are the external resources your assistant can access - email systems, databases, APIs, and other business applications. Outputs define how your assistant presents their completed work - formatted responses, structured data, or other deliverables.

​
Building Your Assistant Step-by-Step

​
Step 1: Set Up Your Assistant’s Identity (Profile)

The Profile is the foundation of your assistant. Start here to establish their core identity and enable collaboration with other assistants.

​
Fill in the Identity Fields

  • Agent Name: Give your assistant a clear, recognizable name (e.g., “Customer Support Assistant”, “Marketing Helper”)
  • Role: Define what job they perform (e.g., “Customer Service Representative”, “Content Creator”)
  • Agent Handle: Create a unique handle like @support, @marketing, or @sales that allows users and other assistants to mention and collaborate with this assistant. This handle enables multi-assistant workflows where assistants can reference each other’s expertise.
Why handles matter: When you create multiple assistants, they can work together by mentioning each other using their handles. For example, a customer service assistant (@support) might bring in a technical specialist (@tech) for complex issues, or a content creator (@marketing) might consult with a data analyst (@analytics) for campaign insights.

​
Write the Description

Define your assistant’s personality and approach to work:
  • Their expertise and responsibilities
  • How they should communicate (formal/casual, brief/detailed)
  • Their approach to problem-solving
Example descriptions:
  • “You are a friendly customer support specialist who helps solve technical issues with patience and clear explanations”
  • “You are a data analyst who provides accurate insights and explains complex findings in simple terms”

​
Set Goals and Constraints

Goals - What should your assistant accomplish:
  • “Resolve customer issues quickly and professionally”
  • “Create engaging content that drives user engagement”
Constraints - What they should NOT do:
  • “Never share confidential customer information”
  • “Escalate billing issues over $500 to the finance team”
  • “Always maintain a professional tone”

​
Configure Knowledge and Model Settings

  • Turn on Embed Knowledge to connect to relevant knowledge bases
  • Choose your AI Provider (OpenAI, Anthropic, Google)
  • Select the appropriate Model for your needs
  • Adjust Creativity Level from Focused to Experimental

​
Step 2: Define How Information Enters (Inputs)

Inputs determine how users will interact with your assistant.

​
Choose Input Types

Select based on what information users need to provide:
  • Text Input: For questions, instructions, or written content
  • File Input: For document uploads (PDFs, Word docs, etc.)
  • URL Input: For web links to analyze
  • Image Input: For photos or images
  • Data Input: For structured data like forms

​
Configure the Prompt

Write a clear description of what type of input you expect: Good prompts:
  • “Customer support questions or issues requiring assistance”
  • “Resume documents for analysis”
  • “Product images for description generation”
Avoid vague prompts:
  • “Input”
  • “Information”
  • “Data”

​
Step 3: Configure Processing Tasks (Activities)

Activities define the core work your assistant performs. You’ll first choose from pre-configured activity types, then write prompts that guide HOW to perform each activity.

​
Choose Activity Types

Select from pre-configured options like:
  • Answer Questions: Respond to user inquiries
  • Summarize Content: Condense information into key points
  • Sentiment Analysis: Analyze emotional tone in text
  • Write Emails: Create email content based on context
  • Custom: Define specialized tasks unique to your needs

​
Write Custom Prompts

Your prompts guide HOW the assistant should perform the chosen activity type: Example prompts:
  • “Analyze the customer support ticket and categorize it as: technical, billing, general inquiry, or complaint. Provide a brief summary.”
  • “Extract the customer’s name, email, order number, and issue description”
  • “Generate a professional email response with an empathetic tone”
Good prompt structure:
  • Start with action words: “Analyze”, “Extract”, “Generate”, “Summarize”
  • Be specific about requirements
  • Include format guidelines
  • Mention constraints

​
Choose Response Type

  • Text: For natural language responses, explanations, summaries
  • Structured (JSON): For data extraction, categorization, consistent data format
When you select Structured, define JSON fields:
  • Field Name: The key in your output (e.g., “priority”, “category”)
  • Description: What this field should contain

​
Select the Right Model

  • GPT-4: Best for complex reasoning and analysis
  • Claude: Excellent for writing and communication
  • Gemini: Good for general-purpose tasks

​
Step 4: Connect External Tools (Tools)

Tools let your assistant take actions beyond processing information.

​
Select Your Tools

Choose based on what actions your assistant needs:
  • Email & Communication: Gmail, Outlook, Teams
  • Project Management: Asana, Linear, Monday.com
  • Customer Management: HubSpot, Salesforce, Attio
  • Custom Integrations: Your organization’s systems

​
Set Up Authentication

  • API Key: Enter your credentials for services using this method
  • OAuth: Use secure flows for platforms like Gmail (no password sharing)
  • Custom: Configure authentication for specialized systems

​
Configure Permissions

Choose the right permission level based on what the tool will do: Human Approval - Recommended for actions that make changes to external systems:
  • Sending emails or messages to customers
  • Creating new records in CRM systems
  • Updating customer data or profiles
  • Creating tasks or projects in project management tools
  • Making purchases or financial transactions
  • Posting content to social media
  • Deleting or modifying existing data
Auto Execute - Suitable for actions that only fetch or read data:
  • Looking up customer information
  • Searching knowledge bases
  • Retrieving project status
  • Checking inventory levels
  • Reading email or message history
  • Generating reports from existing data
Why this matters: Requiring human approval for data retrieval can significantly slow down user interactions and hurt the user experience. Users expect quick responses when asking for information, but they appreciate confirmation before the assistant takes actions that could affect external systems or data.

​
Add Custom Instructions

Define exactly how tools should behave: Example for Email Tool: 
When sending customer support emails:
- Include the ticket reference number in the subject
- Use our customer support signature template
- Copy the support manager on all emails
- Set priority to "High" for urgent issues
Example for CRM Tool: 
When creating customer records:
- Always include lead source in the description
- Set follow-up reminder for 3 days
- Tag based on company size: startup, SMB, or enterprise

​
Step 5: Format Your Outputs (Outputs)

Outputs define the format of results your assistant delivers. Simply choose the format and your assistant will respond accordingly.

​
Choose Output Format

  • Text: For conversational responses, explanations, summaries, and natural language content
  • Structured (JSON): For data extraction, categorization, and information that needs further processing

​
Adding and Connecting Components

​
Adding Components

Simply click and drag any component from the left sidebar onto the main canvas. Each component represents a step in your workflow.

​
Connecting Components

Connect components by dragging lines between them to create your workflow:
  1. Information flows through Inputs
  2. Gets processed by Activities
  3. May use Tools for external actions
  4. Results are delivered through Outputs
  5. Everything operates within Profile guidelines

​
Example Workflow

Customer Support Assistant:
  • Profile: Helpful, professional customer service specialist
  • Input: Customer inquiry via text
  • Activity: Analyze issue and craft response
  • Tool: Check customer history in CRM
  • Output: Formatted email response

​
Finalizing Your Workflow

​
Test Your Assistant

Use the Run button to test your workflow with sample inputs:
  • Try different types of questions or data
  • Check that responses match your expectations
  • Verify tool integrations work correctly
  • Ensure outputs are properly formatted

​
Save and Publish

Once your workflow produces the correct results:
  1. Save your workflow using the toolbar
  2. Publish it to make it available for use
  3. Your assistant will now be accessible from the home page

​
Best Practices Summary

​
Start with a Clear Profile

  • Be specific about your assistant’s role and expertise
  • Set clear boundaries and constraints
  • Choose the right AI model for your needs

​
Design User-Friendly Inputs

  • Make it obvious what users should provide
  • Use clear, specific prompts
  • Choose input types that match how users naturally want to interact

​
Write Effective Activity Prompts

  • Be specific about what you want
  • Include examples when helpful
  • Break complex tasks into simpler steps

​
Configure Tools Securely

  • Use minimum necessary permissions
  • Require human approval for actions that change external systems (mutations)
  • Use auto-execute for data retrieval to maintain good user experience
  • Write clear custom instructions

​
Choose the Right Output Format

  • Use Text for conversational responses and human-readable content
  • Use Structured (JSON) when you need consistent data format for further processing

​
What’s Next?

After building your assistant, you’ll want to:
  1. Deploy Your Assistant: Learn how to make it available to users
  2. Interact with Your Assistant: Understand how to use and manage your deployed assistant
  3. Multi-Assistant Collaboration: Learn how to set up assistants that work together
  4. API Integration: Integrate your assistant with external systems and applications
Your assistant is now ready to help automate workflows and provide value to your users while maintaining the professional standards and constraints you’ve established.
⌘I