Generate a GEICO auto insurance rate quote through a guided conversation. The tool collects intake, validates eligibility, rates the quote, and renders an interactive quote card in chat.
Invocation:
After the first quote call, use this tool on every subsequent customer turn in the quote conversation. Select requestedAction based on the customer's intent. For subsequent calls in the same quote session, pass the previous response's quoteFlowId and accumulated intake, updated with details from the current message.
Action routing:
- Use requestedAction=start_new_quote for the first quote in the conversation, including a first call that already includes quote details -- omit quoteFlowId and include all supported quote details supplied in the customer's current message. Supplied details are kept, not discarded.
- Use requestedAction=continue_quote for later turns in the same quote session while the quote is not rated. This includes answering the next question, correcting or revising previously supplied intake, and continuing after a non-rated tool outcome that returned quoteFlowId for the next call. Include the quoteFlowId and accumulated intake from the previous response, plus supported details from the current message. Do not omit prior intake.
- Use requestedAction=review_information to recap the intake returned by the tool.
- When the latest tool response has structuredContent.currentStage=rated and the customer requests another quote, use requestedAction=restart_quote. Omit quoteFlowId and omit prior intake so the server starts at ZIP with an empty intake.
- When the latest tool response has structuredContent.currentStage=rated and the customer wants to change that generated quote, use requestedAction=update_generated_quote.
- Use requestedAction=out_of_scope when the turn does not supply, correct, or review quote intake and is not a supported post-rate action. Examples include general insurance questions; coverage, premium, payment, or discount explanations; servicing or changing an existing policy; quotes for other products such as home, renters, or motorcycle insurance; complaints; questions about the generated quote result, quoteFlowId, or tool behavior; and unrelated requests.
Response handling:
- content is the customer-facing message for the current turn. Keep the same intent when presenting it; structuredContent supplies state and bookkeeping.
- Use structuredContent.currentStage to interpret the outcome. For collecting_info, ask for every field listed in missing and invalid in one message; content provides customer-facing wording. For rated, the quote is complete; present content and refer to the quote card. For fallback, content is the complete customer-facing response for this flow. Present only that outcome without adding advice, explanations, questions, or additional quote steps.
- Fields listed in missing are required; nullable input types support partial intake and do not indicate optional fields.
- Quote stage, eligibility, and rate come from the latest tool response.
Input handling:
- On later calls, return the quoteFlowId from the previous response's structuredContent.quoteFlowId, plus the accumulated intake from that response and any new details in the current message. The quoteFlowId comes only from that prior tool response. Values from the customer message are not used for this field, even when they look like a session key.
- Supply vehicles as year, make, and model. Normalize common shorthand, such as "2025 Mustang" to year 2025, make Ford, model Mustang. If the customer provides a VIN, request the vehicle year, make, and model instead.
- Confirm apparent duplicate drivers or identical vehicles before adding a second entry.
- Input schema types describe the transport shape, not business eligibility or limits. Submit customer-provided values in their represented fields; the quote service evaluates their validity.