AI Chat Booking Automation Integration
Budget / Salary£750–1,500
TypeFreelance project
LocationRemote
Posted2 hours ago
I want to integrate a production-ready AI conversational booking assistant into my existing UK courier platform. Customers should be able to complete an end-to-end quote and booking through a natural real-time chat interface without using the traditional booking form.
The existing platform already contains the booking flow, pricing engine, address lookup and the relevant APIs. The objective is not to rebuild this functionality, but to add an intelligent conversational layer on top of it.
The assistant should understand natural customer requests, extract information already provided, maintain conversation state and ask only for genuinely missing information.
For example:
“I need a Luton tomorrow from Birmingham to Manchester with a stop in Stoke.”
The assistant should understand the vehicle, date, route and intermediate stop, then continue naturally by requesting only the information still required.
The AI must use controlled functions/API calls for address validation, quotation/pricing and booking creation. Existing platform APIs and business rules must remain the source of truth. The LLM must not calculate prices, invent addresses or reproduce business logic itself, and it should not have unrestricted database access.
The core flow is:
Natural conversation → collect/retain booking information → validate addresses → call existing quotation API → present price → customer confirmation → call existing booking API → return confirmed booking/tracking reference.
The assistant should also provide grounded answers to approved LDC Courier service questions and support human handoff when it cannot confidently complete a request.
Context management is particularly important. Customers must be able to correct information during the conversation, and the assistant must retain valid information already supplied rather than repeatedly asking the same questions.
Multi-stop bookings must be supported.
We also require conversation/error logging and basic admin visibility for monitoring AI performance and failed interactions.
The chat widget must work correctly on desktop and mobile and match the existing platform branding.
Document extraction, OCR and PDF/image processing are explicitly NOT included in this project.
Acceptance criteria
A customer must be able to open the chat, naturally provide shipment requirements, select/validate the required addresses, receive the same price produced by the existing platform pricing engine, confirm the booking and receive a booking/tracking reference. The resulting booking must immediately appear correctly in the existing admin platform.
The implementation should be delivered in small testable milestones. Each milestone must be deployed, demonstrated and available for our testing and acceptance before proceeding to the next major functionality.
Please include in your proposal your recommended architecture, LLM/provider, approach to conversation state and function/tool calling, estimated hours/cost, delivery time, and examples of similar production AI integrations with transactional APIs.
The existing platform already contains the booking flow, pricing engine, address lookup and the relevant APIs. The objective is not to rebuild this functionality, but to add an intelligent conversational layer on top of it.
The assistant should understand natural customer requests, extract information already provided, maintain conversation state and ask only for genuinely missing information.
For example:
“I need a Luton tomorrow from Birmingham to Manchester with a stop in Stoke.”
The assistant should understand the vehicle, date, route and intermediate stop, then continue naturally by requesting only the information still required.
The AI must use controlled functions/API calls for address validation, quotation/pricing and booking creation. Existing platform APIs and business rules must remain the source of truth. The LLM must not calculate prices, invent addresses or reproduce business logic itself, and it should not have unrestricted database access.
The core flow is:
Natural conversation → collect/retain booking information → validate addresses → call existing quotation API → present price → customer confirmation → call existing booking API → return confirmed booking/tracking reference.
The assistant should also provide grounded answers to approved LDC Courier service questions and support human handoff when it cannot confidently complete a request.
Context management is particularly important. Customers must be able to correct information during the conversation, and the assistant must retain valid information already supplied rather than repeatedly asking the same questions.
Multi-stop bookings must be supported.
We also require conversation/error logging and basic admin visibility for monitoring AI performance and failed interactions.
The chat widget must work correctly on desktop and mobile and match the existing platform branding.
Document extraction, OCR and PDF/image processing are explicitly NOT included in this project.
Acceptance criteria
A customer must be able to open the chat, naturally provide shipment requirements, select/validate the required addresses, receive the same price produced by the existing platform pricing engine, confirm the booking and receive a booking/tracking reference. The resulting booking must immediately appear correctly in the existing admin platform.
The implementation should be delivered in small testable milestones. Each milestone must be deployed, demonstrated and available for our testing and acceptance before proceeding to the next major functionality.
Please include in your proposal your recommended architecture, LLM/provider, approach to conversation state and function/tool calling, estimated hours/cost, delivery time, and examples of similar production AI integrations with transactional APIs.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.