Fable
An AI communication coach for hard conversations
An AI communication coach for practising difficult conversations. Work through scenarios, try a different approach, and get feedback on your delivery.
Start with the real need
Difficult conversations are often hard to practise. A blank chatbot gives the user no stakes, no consistent counterpart, and no useful way to reflect on what happened. I wanted to see whether a structured simulation could make rehearsal feel specific enough to be useful.
How I shaped the product
Fable gives each session a scenario, a role for the simulated counterpart, and a clear objective for the user. The model stays inside that situation during the conversation and then produces feedback. The product wraps the model in a deliberate experience instead of treating an open chat box as the product.
Decisions that mattered
Start with a situation
A named relationship, source of tension, and desired outcome give the conversation something concrete to push against.
Separate simulation from reflection
The counterpart should behave like a person in the moment. Feedback belongs after the exchange, when the user can step out of the scene and examine it.
Avoid pretending the score is objective
Model feedback is a prompt for reflection, not a validated assessment of communication skill. The interface and language need to preserve that distinction.
What this does not prove
- The feedback rubric has not been validated against expert coaching or real-world outcomes.
- A simulated counterpart cannot reproduce the history, emotion, or consequences of a real relationship.
- Users may give undue weight to confident feedback produced by the model.
How I would strengthen the evidence
- 01Compare feedback across repeated runs of the same conversation.
- 02Ask communication practitioners to critique the feedback rubric.
- 03Test whether users can name a concrete change they would make in the real conversation.
Interested in the reasoning behind the build?
I share the decisions, failed assumptions, and useful lessons as the work develops.