The TwoStageFallbackPolicy is not yet part of an official release, but you can already try it out as part of the alpha version
The TwoStageFallbackPolicy handles ambiguous user input (classification with confidence below a certain interval) in multiple stages compared to the default fallback policy which can simply ask the user to rephrase.
If the NLU classification confidence is below a given threshold, the TwoStageFallbackPolicy first asks the user to affirm an intent (you can e.g. give him a list of buttons with intents to select from). If the user then affirms the story continues. If the user does not affirm one of the provided intents, they can still rephrase.
Best checkout the documentation of it here: Training and Policies
Please let me know if you have any more questions.