Keeping the tour buses moving

A US luxury tour bus operator manages vehicles and drivers for touring entertainers across the country. Schedules change quickly, bookings evolve constantly and the operational team needs an accurate view of what is happening at any moment.

For more than fifteen years, the business had relied on a custom system that had become increasingly difficult to maintain and extend. It held critical operational knowledge, but years of additions and workarounds made change slower and created growing dependence on a small number of people who understood how everything fitted together.

Replacing it was not a conventional software rollout. The business could not pause while a new platform was built. The new system needed to take shape around a live operation, and the engineering team needed a way to keep shipping complex product work at pace without burning people out or losing the judgment that keeps a live operation safe.

We joined as a fractional product and engineering partner to help design and build Bravo, the custom platform that became the operational system of record. Agentic AI software engineering is central to how that work gets done.

Product sits close to the operation in the US. Development runs remotely from Australia. Shared hours are for decisions: clarifying intent, capturing edge cases and locking the plan while the detail is still fresh. Outside that window, Claude Code works from short written briefs and the engineering rules already set for the codebase, turning those plans into branches, tests and pull requests overnight.

The agents are not left unsupervised. Humans still own architecture, review, merge decisions and anything that touches sensitive paths, especially production data. Feature flags and automated tests are part of the loop, so new work can land in a real environment and be caught before it reaches an operator.

That model made it possible to rebuild the core of the business as one connected platform, including:

• Booking and trip management
• Vehicle availability and tracking
• Driver and operational coordination
• Financial workflows and accounting integration
• Shared operational data and reporting
• Permissions and controls for different teams

A recent feature, Amendment Quotes, shows the pattern. The idea changed substantially during planning, before much code was invested. The revised version shipped behind a feature flag. An automated test caught an edge case involving a declined amendment and a new one starting at the same time. It was fixed before a user ever saw it.

The result is close to continuous coverage for the product: decisions land in the US afternoon, agentic engineering picks them up in Australia, and Bravo keeps moving without anyone working unreasonable hours.

Bravo replaced a fifteen-year-old operational system without pausing the business it supported, and without treating AI as a demo bolted onto a slow delivery process.

The engagement has:

• Made agentic AI part of the default engineering loop, not a side experiment
• Turned written product intent into overnight implementation with human review gates
• Extended the product day across US and Australian time zones without forcing late nights
• Brought bookings, vehicles and finance into one connected system
• Reduced reliance on legacy knowledge and individual workarounds
• Established a tested, documented foundation for continued product development

Most importantly, the business moved onto a new platform while continuing to serve clients, manage tours and keep its buses moving across the United States.

Bravo is live and supporting the operation day to day, but the product is not standing still. We continue as part of the distributed team, using the same agentic delivery model to add features, refine workflows and respond to new operational needs as they emerge. The combination of clear product decisions, Claude Code and human controls lets Bravo evolve at the pace of the business it supports.