An agent team that runs my car-rental business
I built a team of agents that runs the back office of a car-rental business I own. The agents handle day-to-day operations and the work that scales the team, so the business runs largely on the system rather than on me.
The same discipline as the agent work, against real money and real liability: agents do the operational legwork, guardrails govern what can act on its own, and a human stays on every call that carries the risk. It reduces my hands-on time and operating costs, and it is the operator credential that makes the guardrails philosophy concrete rather than theoretical.
The problem
Running a car-rental fleet is a real-time operations problem with money on every decision. Bookings, turnaround between trips, pricing, vehicle condition, and the back-and-forth with each renter all move continuously, and a hands-on owner becomes the bottleneck the moment the fleet grows. The need was the same one the agent work answers: build a system that runs the day-to-day operations and the work that scales the team, so the business depends less on me being present, and keep a human firmly on anything that moves money or cannot be undone.
Approach
I built a team of agents that runs the back office. They cover the recurring operational work end to end, draw on live vehicle telematics as the signal, and surface turnaround and pricing moves off that data rather than off guesswork. The system is bounded by design: agents handle the legwork and the scaling work, and the human approves anything that carries money or is irreversible. It is the same capability-then-leash pattern as the agent work, applied to a real business with physical assets and real liability instead of a software-only action.
- Let agents run the operations. The agent team carries the day-to-day back office and the workload that scales the fleet, working off live vehicle telematics as the signal. The leash: agents do the legwork, they do not own the calls that carry liability.
- Advise, do not transact. The system recommends the turnaround and pricing move the data supports rather than executing it unattended. The leash: a recommendation is not an action until the operator approves it.
- Keep the human on the money. Routine operational work is automated; consequential moves are not. The leash: anything that moves money or is irreversible waits for the operator.
Guardrails
What the system is structurally not allowed to do. This is the through-line: capability, then leash.
- NoThe agents cannot move money on their own. The human approves anything financial.
- NoThe agents cannot take an irreversible action unattended. Those calls wait for the operator.
- NoThe agents do not own a liability decision. The system surfaces the call and does the legwork; the human owns the risk.
My role
I own the business and built the agent team that runs it. It is my own venture, the cars are my own property, and it is where the guardrails philosophy meets real money and real liability. I stay on the decisions that carry liability; the agents carry the rest.
Links and verification
This is a private business, my own venture, so there is no public repo to link and no public metric to cite. It is included as the operator credential behind the guardrails philosophy: a real business run largely by an agent team I built, applied against real money, with a human on the calls that carry liability.