AI Product

When a Deutsche Bahn disruption breaks an onward connection, a booking confirmation is not the end of the customer's problem. This prototype turns booking context, live operations and passenger policy into a clear next step: a deterministic rule owns the money-sensitive missed-connection decision, an LLM handles evidence and interpretation, and anything risky waits for consent or a person. Strategically, post-purchase is the moment that decides whether a traveller trusts the platform with the next journey.

A contextual chatbot designed to complete disrupted multi-leg journeys-not merely answer questions. It combines booking context, live journey state and carrier rules, grounds every recommendation in evidence, exposes uncertainty instead of guessing, and expands from advice to actions only when evaluation proves it is safe.

For an AI finance app's travel insurance: when a covered flight is delayed, this turns the traveller's own policy, live flight data and a fixed payout rule into a clear answer and a pre-filled claim. Coverage and payout amounts are decided by versioned rules, never the AI model; the model only reads the question and explains the result, and a person approves every payout before money moves.