Coworker offers two ways to interact with AI agents: Deep Work for general-purpose tasks and custom Agents for specialized, domain-specific work.
Deep Work
Deep Work is Coworker's flexible, all-purpose AI agent that can handle a wide variety of tasks.
Key Characteristics:
General Purpose: Designed to handle any type of request or task
Flexible Tooling: Has access to all available tools and integrations
Universal System Prompt: Uses comprehensive instructions that work across all domains
Adaptive Planning: Creates execution plans based on the specific task requirements
Broad Knowledge: Can work across sales, engineering, operations, customer success, and other functions
Best Use Cases:
One-off tasks that don't fit a specific category
Cross-functional analysis requiring multiple domains
Exploratory research and investigation
Custom workflows that haven't been standardized
Tasks requiring maximum flexibility and tool access
Agents
Agents are specialized tools designed and optimized for specific business functions, roles, or recurring task types.
Key Characteristics:
Domain Expertise: Specialized knowledge and instructions for specific functions
Optimized Tooling: Pre-configured with the most relevant tools for their domain
Specialized System Prompts: Tailored instructions that understand domain-specific context and requirements
Consistent Output: Standardized approaches and formatting for predictable results
Role-Based Focus: Designed around specific job functions or business processes
Types of Custom Agents:
Sales Agents: Call analysis, pipeline reviews, deal risk assessment
Engineering Agents: Code research, PR analysis, technical documentation
Customer Success Agents: Customer history analysis, health scoring
Operations Agents: Process documentation, workflow optimization
Leadership Agents: Team performance analysis, strategic insights
Key Differences
Aspect | Deep Work | Agents |
Specialization | General-purpose, handles any task | Domain-specific expertise and focus |
System Prompts | Universal instructions for all scenarios | Specialized prompts for specific functions |
Tool Configuration | Access to all available tools | Pre-configured with relevant tools |
Output Consistency | Variable based on task requirements | Standardized format and approach |
Learning Curve | Generally requires more task specification | Understands domain context automatically |
Efficiency | May need more guidance and iteration | Optimized for specific use cases |
Flexibility | Maximum flexibility for any request | Focused efficiency within domain |
When to Use Deep Work
Novel Tasks: When you need something that doesn't fit existing agent categories
Cross-Functional Work: Tasks requiring expertise from multiple domains
Experimentation: Trying new approaches or exploring possibilities
Complex Integration: When you need access to all available tools and data sources
Custom Workflows: Building unique processes that haven't been standardized
When to Use Custom Agents
Recurring Tasks: Regular work that follows predictable patterns
Domain Expertise: When you need specialized knowledge in sales, engineering, etc.
Consistent Output: When you want standardized results and formatting
Efficiency: For faster execution of common business processes
Role-Specific Work: Tasks that align with specific job functions
Examples in Practice
Use Deep Work for:
"Analyze our Q3 performance across sales, engineering, and customer success"
"Create a comprehensive onboarding plan for our new AI initiative"
"Research and compare three different project management approaches"
Use Custom Sales Agent for:
"Analyze my sales calls from last week and identify improvement opportunities"
"Generate a pipeline review for the enterprise segment"
"Assess deal risk for opportunities closing this quarter"
Use Custom Engineering Agent for:
"Review recent pull requests and identify code quality issues"
"Generate technical documentation for the new API endpoints"
"Analyze our GitHub activity and team productivity metrics"
Getting Started
Start with Custom Agents: If your task fits a specific business function, try the relevant custom agent first
Fall Back to HO2: Use the generic prompt bar when custom agents don't meet your needs
Experiment: Try both approaches to see which works better for your specific use cases
Provide Context: Regardless of which you choose, clear context and requirements improve results