diff --git a/design_drafts/refactor_controller_manager.md b/design_drafts/refactor_controller_manager.md
new file mode 100644
index 0000000..d8e8cc8
--- /dev/null
+++ b/design_drafts/refactor_controller_manager.md
@@ -0,0 +1,28 @@
+extract ros2 pub/sub/services into a `ros2_communication` helper files. 
+
+## Motivation:
+In a project, we're wrapping `controller_manager` in a custom scheduler task, essentially re-writing the `ros2_control_node`
+
+- **Conceptual ambiguity**: `ros2_control_node` is not a node, but an executable specific to the OS scheduler responsible for instantiating and running the `controller_manager`, which IS the node.
+- **Code Maintainability**: `controller_manager.cpp` file is hitting 4k LoC, mixes core logic and ROS2 communicaiton (services, publishers, subscribers)
+
+## Proposed changes
+We propose a simple refactoring with two main objectives:
+
+1. **Extract all ROS 2 interfaces** and their callback implementations into a helper file `controller_manager_ros.cpp/hpp` which current `controller_manager.hpp` would include.
+
+2. **Rename the current `ros2_control_node`** to `ros2_control_executable` to accurately reflect its function as a system entry point rather than a Node
+
+## Notes:
+naming is up for discussion, of course
+
+#### Food for thought:
+Outside of scope for this suggestion, but we could push it a step further and make `ControllerManager` not inherit from `Node` at all, and place all ROS-related stuff into `ControllerManagerNode` class.
+
+Then, as we do with `ResourceManager`, we'd pass in the necessary node interfaces to it:
+- executor (already passed)
+- parameters
+- logger
+- clock
+
+Then the `ControlNode` would be responsible for ros2 communication, controller manager for controllers and resource manager for hardware.
\ No newline at end of file