What is Autopilot?
Autopilot is an advanced feature that allows TenderB to autonomously execute complex, multi-step workflows with minimal human intervention. Instead of guiding the AI step-by-step, you define the end goal and Autopilot determines and executes the necessary steps.
Autonomously Switching Between Planning and Agent Mode
A key strength of Autopilot is its ability to switch between Planning Mode and Agent Mode as needed:
Planning Mode: The AI first creates a structured plan or checklist of tasks required to achieve the goal. This ensures a logical workflow and allows for human review before execution.
Agent Mode: The AI then executes the tasks in the plan, generating content, conducting research, and performing analysis.
This autonomous switching ensures that complex requests are handled in a structured, predictable, and verifiable way.
Model Selection in Autopilot
Select the desired model to run Autopilot with.
When to Use Autopilot
Autopilot is ideal for:
Complex tender analysis requiring multiple research steps
Multi-section document generation
Comprehensive requirement extraction and analysis
Market research and competitive analysis
Quality assurance and compliance checking
End-to-end bid preparation workflows
How Autopilot Works
Step 1: Define the Goal
Provide a high-level objective rather than step-by-step instructions.
Example:
"Please create a complete tender response addressing all requirements in the tender document. The response should include: executive summary, technical approach, team and experience, pricing strategy, and implementation plan. Ensure all mandatory requirements are addressed and content is verified against source documents."
Step 2: Enable Autopilot Mode
Select "Agent Mode" which includes autopilot capabilities.
Step 3: AI Switches to Planning Mode
Autopilot first enters Planning Mode to create a fixed checklist. It will present a structured plan, often in phases, with a list of to-do items. You'll see a message like "Created 12 to-dos" indicating the plan has been generated.
Step 4: Review and Approve the Plan
Click "View results" or "Expand plan" to see the detailed checklist. Review the plan to ensure it aligns with your expectations. The plan typically includes:
Multiple phases (e.g., Phase 1: Analysis & Research, Phase 2: Content Generation, etc.)
Specific tasks within each phase
Key focus points for each task
A status indicator showing which phase is ready to begin
You can ask for modifications to the plan before execution begins.
Step 5: AI Switches to Agent Mode for Execution
Once the plan is approved (or you respond "yes" to proceed), Autopilot switches to Agent Mode to execute the tasks. It provides updates as it progresses through the checklist. You'll see the AI working through each phase systematically.
Step 6: Review Output
Autopilot presents completed work with explanations of decisions made.
Step 7: Refine if Needed
Request specific improvements: "Please strengthen the risk management section with more specific mitigation strategies."
Autopilot Capabilities
Autonomous Planning:
Breaks down complex goals into steps
Determines optimal sequence
Identifies required information
Adjusts plan based on findings
Multi-Agent Orchestration:
Automatically engages appropriate sub-agents
Coordinates parallel work streams
Synthesizes outputs from multiple agents
Ensures consistency across sections
Research and Analysis:
Conducts web research when needed
Analyzes multiple documents
Extracts and synthesizes information
Identifies gaps and inconsistencies
Quality Assurance:
Automatically verifies content against sources
Checks for completeness
Ensures compliance with requirements
Flags potential issues
Monitoring and Control
Progress Visibility:
Real-time updates on current step
Ability to view the plan at any time by clicking "Expand plan"
Ability to view intermediate results
Status indicators showing which phase is active
Intervention Options:
Pause autopilot to review progress
Provide additional guidance mid-process
Adjust direction based on findings
Stop and restart with refined instructions
Best Practices for Autopilot
Clear End Goal: Define what success looks like
Complete Resources: Provide all necessary documents upfront
Appropriate Scope: Don't make goals too broad or too narrow
Review the Plan: Always review the plan before execution to ensure it's on track
Trust but Verify: Review autopilot output carefully
Iterative Refinement: Use autopilot results as starting point for refinement
Learn from Results: Note what works well to improve future requests
Autopilot vs. Manual Mode
Use Autopilot When:
You have a clear end goal but don't want to specify every step
The task requires multiple research or analysis steps
You need parallel processing of multiple sections
Time is limited and you need efficient execution
Use Manual Mode When:
You want precise control over each step
The task is simple and straightforward
You're learning how to use TenderB effectively
You need to make decisions at each step
Related Articles:

