Two Core Modes of Operation
TenderB operates in two primary modes: Planning Mode and Agent Mode. The AI can switch between these modes autonomously to ensure a structured and effective workflow.
Planning Mode: Structuring Your Work
Purpose: To create a structured plan or checklist before executing a complex task.
In Planning Mode, the AI does not generate content directly. Instead, it analyzes your request and creates a step-by-step plan to achieve your goal. This plan is presented as a fixed checklist or a series of phases.
What You'll See in Planning Mode:
A message like "Created 12 to-dos" indicating the plan has been generated
A "View results" or "Expand plan" button to see the detailed checklist
Multiple phases with descriptive names (e.g., "Phase 1: Analyse & Research (Foundation)")
Specific tasks and focus points within each phase
A status indicator (e.g., "Status: ✅ Plan gereed. Klaar om aan Fase 1 te beginnen.")
Key Features of Planning Mode:
Creates Checklists: Breaks down large tasks into manageable to-do items.
Defines Phases: Organizes the workflow into logical stages (typically 6 phases for complex bids).
Human Review: Allows you to review and approve the plan before execution.
No Content Generation: The AI focuses solely on planning, not writing.
Fixed Structure: The plan provides a clear, predictable path for the AI to follow.
Transparency: You can see exactly what the AI intends to do before it starts.
Example of a Planning Mode Output:
When you ask TenderB to create a comprehensive bid, it might generate a plan like this:
Plan Overview: Winnende Aanbieding Telefonie BOOR
Fase 1: Analyse & Research (Foundation)
Tender Analyse: Extract all requirements, sub-criteria, format requirements, and evaluation methodology
TenderB Research: Deep dive into knowledge base for company's previous projects, certifications, team expertise
Competitive Context: Research market positioning and differentiators
Fase 2: Strategic Foundation
Win themes identification
Value proposition development
Risk analysis
...and so on through Phase 6: Finalisering (Deployment)
When is Planning Mode Used?
TenderB automatically enters Planning Mode when you:
Use Autopilot for a complex request.
Ask the AI to create a plan for a large task (e.g., "Create a plan to write this entire tender response").
The AI determines that the request is too complex for immediate execution.
Agent Mode: Executing the Plan
Purpose: To execute tasks, generate content, and perform analysis.
In Agent Mode, the AI actively works on the tasks defined in the plan. This is where content is written, research is conducted, and analysis is performed.
What You'll See in Agent Mode:
The AI actively writing content in the chat
Progress updates as it works through the checklist
Sub-agents being engaged for specialized tasks
Content appearing in the "Write" panel on the right side
The ability to switch between "Agent" and "Marlowe" (the main AI assistant)
Key Features of Agent Mode:
Content Generation: Writes text, creates tables, and generates other content.
Task Execution: Follows the checklist created in Planning Mode.
Interactive: You can provide feedback and request refinements.
Uses Sub-Agents: Engages specialized agents (Writing, Research, etc.) to perform tasks.
Real-time Output: You see the work being produced as it happens.
Important Note About Editing:
When in Planning Mode, you'll see a message: "Agent cannot edit in plan mode. Please switch to agent mode if you would like the agent to edit your project."
This means that while the AI is in Planning Mode, it cannot make changes to your document. You need to approve the plan and allow it to switch to Agent Mode before content generation begins.
Switching Between Modes
TenderB handles the switching between Planning and Agent mode autonomously. The typical workflow is:
User Request: You provide a high-level goal.
Switch to Planning Mode: The AI creates a structured plan.
User Approval: You review and approve the plan (or respond "yes").
Switch to Agent Mode: The AI executes the plan step-by-step.
You can also manually influence the mode by the type of request you make. A request to "create a plan" will trigger Planning Mode, while a request to "write a section" will trigger Agent Mode directly.
Why This Matters: Reducing Hallucinations
This dual-mode system is a powerful way to reduce AI hallucinations and ensure high-quality output. By first creating a fixed checklist in Planning Mode, you provide a clear, structured path for the AI to follow. This prevents the AI from deviating from the requirements and inventing information. The plan acts as a set of guardrails, keeping the AI focused on the specific tasks needed to achieve your goal.
Benefits of the Dual-Mode System:
Structure: Ensures a logical and predictable workflow.
Verifiability: The plan serves as a checklist to verify completeness.
Control: Allows for human oversight and approval before execution.
Accuracy: Reduces the risk of AI going off-topic or inventing content.
Efficiency: Breaks down complex tasks into manageable steps.
Transparency: You know exactly what the AI will do before it starts.
Quality: Each phase can be reviewed and refined before moving to the next.
Tips for Working with Both Modes
Always review the plan: Before approving execution, carefully review the phases and tasks.
Ask for plan modifications: If something doesn't look right, ask the AI to adjust the plan before execution.
Use "Expand plan" frequently: Keep the plan visible to track progress during execution.
Understand the mode you're in: If you see "Agent cannot edit in plan mode," you know you need to approve the plan first.
Leverage the structure: Use the phases as milestones to review and refine work incrementally.