Developers can engage a conversational experience when prompting to plan steps and actions before sending the back-end call for rendering.

Benefits of planning mode

Prompting is an iterative design activity. Developers routinely refine intent, sequencing, and constraints as they converge on a correct and maintainable solution. Plan mode supports this activity by exposing what the system will generate before execution, allowing developers to validate and adjust prompts without committing to a render.

By delivering a step-by-step plan of what will be rendered, Plan mode enables early validation of intent. Developers can confirm that the AI has interpreted the prompt correctly, assess step ordering, and identify gaps or incorrect assumptions before committing to a render. This is particularly valuable during early design and ideation phases, where requirements are still evolving and rapid experimentation is expected.

Reviewing and correcting a plan first avoids repeated renders caused by minor prompt errors, thereby improving efficiency and accelerating automation creations.

Collectively, these benefits enable a more conversational, iterative development workflow. Developers can engage AI as a collaborative design partner to achieve high-quality outcomes without sacrificing rigor, governance, or performance expectations.

Activating planning mode

Note: Build mode is the default operating mode of Co-Pilot for Automators. In this mode, approved prompts result in back-end calls that generate automation artifacts. Each step is processed per request, and upon completion the system remains in Build mode for subsequent interactions.
Activate planning mode by typing /plan in the Co-Pilot interface. All subsequent prompts will be processed using the planning mode, until deactivation. This model separates design from execution, enabling controlled iteration without altering the default build workflow. See the following examples:
  • /plan a task to send emails to all contacts in spreadsheet $Client List$.
  • /plan a process that requests manager approval before a PTO request is accepted.

Planning Mode activates a planning agent that analyzes a prompt without rendering automation. The agent translates the prompt into an ordered set of steps that represent the automation flow.

Key characteristics:

  • Pre-render analysis only; no automation artifacts are created.

  • Step-level output describing structure, sequencing, and dependencies.

  • Agentic, multi-turn interaction; the agent might request clarification when intent or constraints are unclear.

  • Low-friction iteration; prompts can be edited and re-evaluated repeatedly before execution.

The output is a summarized automation plan intended for review and refinement.

Deactivate planning mode

You can easily deactivate planning mode using either of two methods.
  • After completing your plan, simply type build it to create the automation artifact and deactivate planning mode.
  • To terminate the planning session without building your plan, simply close the interface. Your next session opens in the build mode by default.