Create an AI Player That Talks, Acts & Uses Game State (Part 2 - Custom Actions)

Hello everyone,

Following up on Part 1, we are excited to release Part 2 of our AI Teammate tutorial series in Unreal Engine! In this continuation, we extend our Stack-O-Bot AI companion with Custom Convai Actions so it can pick up objects, carry them across moving platforms, and place them on pressure plates to solve co-op puzzles.

While Part 1 covered live game state and built-in navigation, Part 2 shows you how to hook custom gameplay logic into Convai’s Action system—giving your AI player physical agency in the game world through voice commands.

:sparkles: What We Cover in Part 2

This step-by-step teaching tutorial guides you through writing custom action handlers from scratch:

  • :package: Custom Convai Actions: Define custom pickup and drop-at actions complete with target and destination parameters.

  • :hammer_and_wrench: Blueprint Action Handlers: Use Actor References for scene objects and connect AI actions directly into Stack-O-Bot’s existing interaction system.

  • :counterclockwise_arrows_button: Success & Failure States: Handle action state responses so the AI character knows whether an interaction succeeded or failed.

  • :puzzle_piece: Coordinated Puzzle Solving: Watch the AI teammate understand player commands, grab a crate, navigate platforms, and place the crate to unlock the next area.

:link: Project Files & Helpful Resources

Follow along using our downloadable project files, setup guides, and prerequisites:

Prerequisite & Workflow Videos:

How are you using custom actions in your games or XR projects? Share your questions or project updates in the replies below!

Happy developing!