Transport

Turning every call into a booking.

A Los Angeles private car service needed phone and web bookings to flow automatically into its dispatch system, with coverage that didn't stop when staff went home. SDTC Digital connected phone calls, web chat, live pricing, and dispatch, then built a path for the operator to resell the same system to other companies.

Channels: Phone and web chat
Books into: Their dispatch system
Cover: 24/7, including after hours
New operator live in: Under 15 minutes

Project overview

A private car and limo operator in Los Angeles, running every reservation through a person. Calls were answered when somebody was free, website chat was answered when somebody noticed, and the details of both were then retyped by hand into LimoAnywhere — the dispatch system the whole business runs on.

The brief was not simply to bolt a chatbot onto the website. It was to take bookings end to end on two channels, land them reliably in the system of record, and then turn the result into something other limo companies could subscribe to.

System of record: LimoAnywhere dispatch
Delivery shape: Core bot → intelligent product → SaaS platform
Ownership: Operator configures the bot without a developer

Project details

Provider or platform: Retell AI + Twilio (voice)
Data standard: Vector embeddings (OpenAI) for the knowledge base
Access model: Operator configures the bot without a developer
Tenancy or scale: Core bot → intelligent product → SaaS platform

The Challenge

01Every booking depended on a human answering the phone or replying to chat, leaving nothing covered after hours.

02Website visitors had no direct way to book, only to express interest.

03Call and chat details were manually re-entered into LimoAnywhere, which is slow and drops information.

04No searchable record of past calls or chats existed, so nothing could be checked after the fact.

05Any change to the bot's behaviour required a developer, keeping the operator at arm's length from his own product.

Project goals

Automate inbound bookings across both phone and website chat.

Push every booking reliably into LimoAnywhere rather than into a parallel system.

Give callers live rate quotes instead of a promise to call back.

Build a knowledge base so the bot can answer service-area and add-on questions itself.

Give the operator an admin panel to configure and monitor the bot.

Lay the groundwork to sell the product to other car and limo companies.

What We Built

SDTC Digital built the booking agent in three phases on one shared architecture, so each phase extended the last rather than replacing it.

Core booking bot

A voice and chat agent that collects a full booking conversationally, writes it directly into LimoAnywhere, and sends an SMS confirmation, shipped with a basic admin panel for setup and monitoring.

Intelligent product

Live rate quotes, knowledge-base answers, call recording and transcription, a richer chat experience with a date picker, quick replies and address autocomplete, email confirmations, after-hours callback handling, and an analytics dashboard.

SaaS platform

Self-service signup and onboarding for new operators, Stripe subscription billing, a super-admin portal spanning every client account, and white-label branding, turning a single-client tool into a sellable multi-tenant product.

The User Journey

Call or start a chat

The customer phones in or opens the website chat to book a ride.

The bot collects the trip

Pickup, drop-off, time, and any add-ons conversationally, with landmarks resolved to real addresses.

The booking is written into LimoAnywhere

The reservation is created directly in the dispatch system, with automatic retry if the system is briefly unavailable.

Confirmation goes out

The customer receives an SMS confirming the booking details.

Questions get answered anytime

The bot answers service-area and add-on questions from the knowledge base, at any hour.

The Thinking Behind the Platform

01Real conversation, not a phone tree: Retell AI over Twilio lets callers interrupt, change their minds and ask questions mid-booking.

02Semantic understanding: The company site and documents are converted to vector embeddings, so questions are matched by meaning instead of keywords.

03Landmarks resolved to addresses: A reference like “the Hollywood Bowl” is confirmed back to the caller as a real address before booking.

04Self-tuning by the operator: Plain-English instructions typed into the admin panel update the bot's behaviour with no developer involved.

05Fails safely: If the dispatch system is slow or down, bookings queue and retry, with escalation to a human when needed.

06Every call searchable: Calls are recorded and transcribed with Whisper, so a disputed booking can be checked against what was said.

Tech stack

Application

NestJS
Node.js
React
Preact

Voice and AI

Retell AI
Twilio
OpenAI Whisper
OpenAI Embeddings

Data and platform

PostgreSQL
pgvector
Stripe
Clerk
Railway
Cloudflare R2

Testing and Delivery

2,740+

Test cases built into the QA process across all three phases.

Isolation scenarios tested in the multi-tenant security suite.

Zero

Cross-tenant data leaks found across those 36 scenarios.

One

Reservation produced per duplicate or concurrent booking attempt, never two.

Outcomes and Business impact

Bookings complete end to end without a human on the line, on both phone and chat.

Failure handling means a booking is never silently lost, even when the dispatch system has a issue.

New operator companies can go from signup to a live bot in under fifteen minutes.

Missed-call reduction, booking-volume increase and time saved are not yet measured; the recommendation is to capture these from the client post-launch rather than estimate them.