Route SixtySixt

A route to agentic upselling

SIXT sales representatives possess skills that are extremely difficult for AI computer systems to replicate. They can assess their customers and instantly identify which add-ons to recommend. They use humour, psychological techniques and tactics that AI systems are currently incapable of. This is where our tool comes in: Our tool enriches contextual data and feeds it into a Gemini-powered agentic AI, which suggests upgrades that make the most sense to the customer in context.

Recommending upgrades that feel right for the customer based on their trip.

Inspiration

Car rental sales at automated pickup points face a critical dilemma: While customers expect lightning-fast convenience (scan QR, grab keys), the current process offers no opportunity for tailored sales guidance. This means companies miss out on critical revenue, and customers often arrive unaware of upcoming weather hazards, the true value of protection plans, or the benefit of an upgraded model for their actual route.

Introducing Route SixtySixt: Sales Without Delay

We built Route SixtySixt to solve this. It’s an intelligent journey analyst that reasons about each traveler's specific trip before they reach the kiosk.

By leveraging booking details and real-time trip data, Route SixtySixt anticipates customer needs. This system enables high-value upsell moments that are precisely timed and genuinely helpful, allowing the rental company to drive significant new revenue without compromising the customer's expectation of a zero-friction pickup experience.

What it does

Route SixtySixt creates personalized sales opportunities without adding friction to the automated pickup experience. The system's core is a Large Language Model (LLM) that acts as an intelligent journey analyst. It works by fusing two key data streams: the traveler's existing Booking Data (like group size and pickup, dropoff context) with dynamic Local/Real-Time Data (such as live weather forecasts and traffic alerts). This fusion allows the LLM to reason about the unique requirements and potential risks of the specific trip. It then identifies the single, most helpful upsell, be it essential protection or a necessary upgrade and delivers it as a targeted, timely notification at the pickup kiosk, ensuring maximum relevance and zero delay.

How we built it

Route SixtySix uses contextual data gathered from the SIXT API, the Google Maps API, the Google Weather API and so on, which is then fed into our agentic Google Gemini backend. We then combine this with an emotion recognition system at the point of collection and detailed historic customer information to recommend the upgrades with the highest chance of being purchased.

Challenges we ran into

Utilizing all the Google API sources required significant effort. In addition, providing the right context and the right system prompt for Gemini so it can accurately interpret contextual trip data and ensuring the delivery of a good nudge - a genuinely helpful recommendation - without being pushy or manipulative. This relied heavily on successfully finding meaningful contextual data by integrating and managing the Google Cloud API suite and third-party services in a low-latency pipeline, a process critical for maintaining the essential zero-friction customer experience.

Accomplishments that we're proud of

We are most proud of the fact that our system provides intelligent, context-based recommendations that are helpful at the pickup station without being pushy. Customers don't want to complete a 20-question survey or spend 30 minutes talking to an AI voice agent. They want a quick collection process with recommendations that are meaningful to them. We believe that this is an area in which we excel.

What we learned

We learned that context is paramount: the success of an agent-based sales system depends entirely on the quality and relevance of the data provided to the AI agent, rather than just the size of the model. Furthermore, achieving the right balance between maximising revenue and making customers feel guided rather than pressured is an ongoing engineering challenge, even with today's AI capabilities.

What's next for Route SixtySixt

We plan to expand the system’s agentic capabilities by integrating post-pickup communication features, allowing for dynamic, in-trip risk mitigation and real-time support based on current vehicle telemetry and evolving trip conditions. This will involve decentralizing specific processes currently handled by the large LLM - such as risk scoring and fuel consumption prediction - into smaller, specialized agents to improve efficiency and latency. Our focus is on making the LLM a comprehensive travel assistant that moves beyond the initial pickup transaction.