For skiers and snowboarders, every moment on the mountain is about maximizing the fun — chasing fresh lines, perfecting a new trick, or exploring new terrain. Whether they’re exploring a familiar favorite or visiting a new mountain for the first time, riders want the right info at their fingertips so they can move with confidence, find hidden gems, and feel like they belong from the moment they arrive.
It’s a big part of the reason Vail Resorts launched My Epic Assistant during the 2024-2025 snow season. Vail operates some of the most iconic and beloved mountain destinations in the world (Whistler Blackcomb, Park City Mountain, Stowe, and Crested Butte, to name a few). The company wanted every guest to feel fully supported by their new app — able to get quick, helpful answers and discover everything their mountain has to offer, all in a way that keeps them immersed in the experience.
Just think, would you rather be looking for an info booth, or getting help from the app while riding the lift to your next run?
My Epic Assistant is an AI-powered assistant that takes the expertise of Vail Resort’s IT, hospitality, and operational teams and feeds it into Google’s powerful Gemini models. The result is an agentic guide to the slopes that can help decide on the right season pass, share the latest snow report, check on lesson preparations, or suggest a good stop for cocoa. Vail Resorts wanted more than a chatbot; they wanted a digital concierge that understands the nuance of a powder day at Whistler versus a family trip to Beaver Creek.
To take these offerings to the next level, the teams at Vail Resorts and Google Cloud specialist 66degrees updated My Epic Assistant with new features for the 2025-2026 season, including season pass recommendations and improved personalization. This feature now intelligently guides guests to the best pass for them — understanding guest questions and handling complex requests.
In this post, we’ll guide you through how we navigated the moguls of orchestrating a multi-agent system that can effectively answer the complex and nuanced questions from discerning customers looking to get help fast so they can get back to enjoying the mountain. It’s not just savvy guests who can benefit from AI-powered tools like these, either.
It’s not just savvy guests who can benefit from AI-powered tools like these, either.
In this blog post, we’ll detail how we built the My Epic Assistant, orchestrating a multi-agent system that can expertly slalom around the unique challenges of this rugged corner of the hospitality industry. It’s an approach that organizations in other customer-centric sectors can follow to help build agentic system for their own unique situations.
A new season, and better resultsAs we’ve implemented and refined personalization, improved search, summary capabilities, and conversational flow within My Epic Assistant, the app has achieved a 45% reduction in escalation to human agents since its initial launch.By automating routine logistics and expanding personalization, Vail Resorts’ Guest Experience Technology team has been able to ensure that technology never replaces the personal touch, it only enhances it. The overall experience for guests seeking support is now faster, more reliable, and available in more ways and times.According to the team, “Utilizing tooling from Google Cloud, we could lean into agentic design patterns that gave us a way to unlock natural, personalized conversations. These boosted customer satisfaction, while reducing the need for manual intent design. These tools also let us combine flexibility and control to enable the assistant to respond fluidly but always within the boundaries of our brand, policies, and product strategy.”
The steps to building My Epic Assistant
First, My Epic Assistant uses a subtopic classification agent to understand the guest’s initial request — whether they want to compare passes, get a recommendation, or just find general information. A date object is used to determine the time of year, as passes are only available for purchase for a limited time. If passes are no longer on sale, it will automatically route to information on lift tickets. If both passes and lift tickets are available, it asks a few clarifying questions to help customers decide which option will provide the best value for their trip.Next, the assistant hands off to the data collection agent, to determine what additional data is needed to provide a recommendation. For authenticated users, Vail Resorts provides their existing data like their home resort, average visits, age, upcoming visit resort, and peak day preferences called from a webhook. For a new non-authenticated user, it asks clarifying, easy to answer questions to fill in the gaps. There’s no one-way, static form filling here!
The generative AI’s conversational capabilities really shine during this step. My Epic Assistant is designed to chat freely back and forth, allowing guests to ask clarifying questions at any point and then return to the recommendation process without losing context. Via the instructions in the playbook, the model continually assesses if all parameters have been filled for the output in order to move on to the next step, and still allows guests to ask other questions during this process.
For instance, it’s possible that when asked if they want peak-day access, a guest might follow up with “What are peak-restricted dates this year?” My Epic Assistant will respond by invoking the website datastores that codify Vail Resorts’ extensive knowledge. My Epic Assistant then provides an answer and returns to the previous turn in conversation without losing context.
Finally, the recommendation agent takes over. Using the collected user data, it queries a structured database of all pass options to find the perfect match. The system then generates a user-friendly response explaining why an identified pass is a good fit and provides content cards with direct purchase links, simplifying the final step for the guest.
The core functionality of the pass recommendation agent is a datastore tool of a pass matrix that contains a structured file of all existing passes, their features, and their restrictions. With a broad number of options available, extensive testing was necessary to ensure the tool returned valid and appropriate options once it received all of the input parameters, and 66degrees relied on Vail Resorts’ deep domain knowledge to validate every possible outcome.
Then, prompt engineering ensures that the ultimate response presents the recommendations with detailed explanations as to why those specific passes were the best options for them. Content cards are also invoked with purchase links for those recommendations, without having to exit the playbook. For some occasions, a generative response was not appropriate, such as with specific details on the newly launched Epic Friends feature. Static responses are called via a code snippet and a conditional action, again housed within the playbook itself, simplifying the overall architecture.
Personalization made easy as a bunny slope
Thanks to the new combination of smart technology and user-focused design in the My Epic Assistant, Vail Resorts moves beyond simple customer Q&A to a truly conversational yet fully automated experience for many customer requests. It’s about providing the same concierge services resort-goers expect from an operator like Vail Resorts, yet achieving them at scale that’s impossible without AI and cloud technologies.
With pass recommendations from My Epic Assistant, Vail Resorts is gearing up every guest to receive custom recommendations for their best season yet.
Want to turn your technical black-diamond challenges into easy green groomers? Contact the experts at 66degrees and Google Cloud to discover what’s possible with the latest AI and cloud technologies.