Our client runs black car and chauffeur operations across Nassau County and into New York City. Before we built this system, their operations ran on group chats and memory. It worked — until it didn't.
The Situation
Quote requests came in by text and call throughout the day. A customer would message asking for a price on an airport run or a corporate transfer. Someone would see it and respond, or not see it and respond late. Response time varied from minutes to hours depending on who was available.
Dispatchers managed active jobs in group chats. When a driver was en route, updates lived in the thread. When a trip finished, follow-up with the customer — confirmation, receipt, feedback request — depended on whoever remembered. There was no system enforcing it. Some customers got follow-ups. Others didn't.
The operation had no single view of what was happening. A dispatcher wanting to check on a job had to scroll back through messages. Context was buried. Coordination worked through repetition and re-asking.
The business was growing. That growth was making the fragility more visible.
What We Built
We deployed a Claude-powered SMS agent into the company's existing phone number flow. The agent handles the routine operational layer — the requests and follow-ups that were consuming dispatcher attention without requiring dispatcher judgment.
Inbound quote requests are handled automatically. The agent collects trip details through a short SMS exchange — pickup location, destination, date, time, vehicle preference — then calculates pricing against the company's rate schedule and sends a formatted quote. Standard requests go from inbound text to quoted price in under two minutes, without a dispatcher touching it.
Confirmed jobs route to the dispatcher queue with full context attached: all trip details, the customer's message history, vehicle type, and any notes the agent collected. Dispatchers see a structured entry, not a group chat thread.
After trip completion, the agent sends follow-up messages on a defined schedule — confirmation of completed service, a prompt for any questions, and a request for feedback. The timing and content follow the company's service standards. No dispatcher needs to remember to send it.
The system runs on the same phone number the company always used. Customers don't know the difference. Dispatchers see a cleaner queue.
Results
Quote response time dropped to under two minutes for standard requests. The previous baseline was hours — not because no one was trying, but because responding to every incoming text manually was not a sustainable first priority against managing active jobs.
Follow-up completion is now 100% on completed trips. The system sends it; it does not depend on memory.
The most useful signal came from the team itself. They created a separate channel for manual and complex reservations because the AI was handling routine volume at a level that warranted its own lane. That kind of internal reorganization around a system is a reliable indicator it is doing its job.
The dispatcher role did not shrink. It changed. Complex reservations, relationship accounts, exception handling — those still require a person. The routine coordination that used to share that attention now does not.
Want results like this? Let's talk →