Plan mode

Plan mode is in closed beta
Plan mode is currently in closed beta. Results may not be accurate or complete. If you find issues or have suggestions, please reach out to us to help us improve PostHog AI.

Plan mode is a structured workflow for tackling complex, multi-step tasks. Instead of executing immediately, the agent first creates a plan for your approval, ensuring you're aligned on the approach before any work begins.

The agent will ask clarifying questions if needed, create a plan document outlining the steps, and wait for your approval before executing. This prevents wasted effort on misunderstood requirements and gives you visibility into how the agent will approach your task.

When to use plan mode

Plan mode is useful for:

  • Multi-step analyses - Why did conversion drop last week? requires checking multiple metrics and correlating data
  • Complex investigations - Tasks that touch multiple data sources or require a systematic approach
  • Dashboard creation - Building comprehensive dashboards with multiple related insights
  • Cross-functional analysis - Questions that span product analytics, session replays, and error tracking

For simple questions like "show me daily active users", the regular chat mode is faster and more appropriate.

How plan mode works

Step 1: Clarification

When you enter plan mode with a request, the agent may ask clarifying questions to understand your objective. These questions appear as an interactive form that you can fill out. For example:

  • What time period should the analysis cover?
  • Which user segments are you interested in?
  • What metrics define success for this analysis?

Step 2: Planning

The agent presents a plan that outlines:

  • The steps it will take
  • What data sources it will use
  • What insights or outputs you'll receive

Step 3: Review

You review the plan and can:

  • Approve - The agent proceeds to execution
  • Request changes - Ask the agent to modify the plan
  • Cancel - Exit plan mode without executing

Step 4: Execution

Once approved, the agent switches to execution mode and works through the plan, creating insights, running queries, and producing the outputs outlined in your approved plan.

Entering plan mode

You can enter plan mode by:

  1. Selecting the "Plan" option in the mode selector above the chat interface
  2. Asking explicitly - "Create a plan for analyzing our onboarding funnel" or "Plan how to investigate our checkout drop-off"

What you can do in plan mode

In plan mode, the agent can:

  • Explore your data schema - Understand what events, properties, and data sources are available
  • Search existing content - Find relevant insights, dashboards, and notebooks
  • Ask clarifying questions - Gather requirements through forms and questions
  • Read documentation - Access PostHog docs to understand best practices
  • Create plan notebooks - PostHog AI creates plan notebooks that are initially transient (visible only in the conversation). Click Create notebook to save it to your project, making it accessible at a permanent URL. Once saved, any updates to the notebook are automatically synced.

What you cannot do in plan mode

Until you approve the plan, the agent cannot:

  • Create or modify insights
  • Create or modify dashboards
  • Run any operations that modify your data or content

This ensures you maintain control over what changes are made in your project.

Plan mode vs regular chat

AspectRegular chatPlan mode
Best forQuick questions, simple tasksComplex, multi-step analyses
PlanningAgent works immediatelyAgent creates a plan first
ApprovalNot requiredRequired before execution
VisibilitySee results as they happenSee plan before execution

Community questions

Was this page useful?

Questions about this page? or post a community question.