Working example / Free, no signup
What if your course
could talk back?
Give learners room to find their own words. Try a conversation with an AI customer, examine the feedback, then take the design and make it yours.
01 / Step into the conversation
The replacement that never arrived.
Fictional practice. AI feedback can miss nuance. Use it to reflect and retry; it is not a validated assessment.
Your focus this attempt:
A conversation, in your own words
Alex is waiting. What would you say?
- Keep names, order numbers, and workplace details fictional. Your replies are sent to OpenAI, or Anthropic if the primary service is unavailable, to generate the conversation and feedback.
- The transcript stays in this tab and is not saved to a Lab account. Provider retention policies apply.
- Live AI usage is limited so this can stay free. The design kit below is always available.
You have an unsent reply. Send it to Alex, or discard it and review your feedback.
This attempt is complete. Your conversation stays here so you can check the evidence.
02 / Look at the evidence
A closer look at your conversation.
Quotes are checked against your replies. The interpretation is AI-generated coaching and may be mistaken. No numerical score or certification is assigned.
03 / Take the design with you
Build the conversation your learners need.
The free kit includes the character instructions, scenario facts, observable feedback criteria, and a guide to adapting the pattern. Copy it, change it, use it in your own work.
Keep the learning decision visible.
- Define the behavior. Decide what learners should practice before writing a character.
- Give the character boundaries. Define facts, constraints, and what changes its response.
- Separate conversation and coaching. The character reacts. The coach examines the attempt.
- Make feedback inspectable. Link observations to learner speech and let people try again.
Using this in a course
You can link to this public demo. For an embedded, customized version, host your own interface and server endpoint, keep provider keys on the server, and set usage limits. The public demo is not a SCORM package and does not report completion to an LMS. The kit explains the integration choices.