How Sandbox VR gives 1,000+ employees instant operational answers across 90 locations with Chatbase
500-1,000
questions resolved per week
15-20%
reduction in onboarding for new employees
93%
resolution accuracy
At a glance
Sandbox VR is the world's premier immersive virtual reality entertainment company, operating 90+ venues globally with proprietary full-body motion capture technology, in-house AAA game development, and a Netflix collaboration. Backed by a16z and Alibaba, the company has surpassed $200M in lifetime sales and was named #2 on Fast Company's Most Innovative Companies list in AR/VR for 2024. With over 1,000 employees across corporate and franchise operations and roughly 30 new locations opened in the past year alone, the team needed a way to get operational answers to frontline staff instantly, in any language, and speed up onboarding as new locations came online. They built Apollo, an internal AI knowledge agent powered by Chatbase, deployed directly in Slack.
“Apollo is essentially our internal knowledge agent where employees can ask and receive answers on daily operations, tech, and service instantly.”
Finding the right SOP took employees longer than fixing the problem
Sandbox VR runs complex, hardware-intensive experiences for paying guests. When a tech issue hits during a live session, the team on the floor needs an answer in seconds. Before Apollo, employees searched the resource center for the right SOP, spending 2 to 5 minutes per question. Or they asked a manager on Slack and waited. At one location, a tech issue caused by user error took the T1 support team 2.5 hours to diagnose. When the same question was put to Apollo, it identified the root cause in seconds, a time saving that multiplies across 90 locations. The team had trained employees on 100% of possible issues, no matter how rare. That made onboarding slow and forced frontline staff to memorize procedures they might use once a year.
“At Sandbox VR, when you need information, you need it very fast. There could be a live tech issue happening in a live experience and if that's happening you need to resolve that issue as fast as possible in order to support those customers.”
2,900 questions in the pilot, 93% accuracy in production
During a 6-week pilot, Apollo handled 2,900+ interactions. Since launching globally, the team sees 500 to 1,000 questions per week across 90 locations. Only around 5 instances have required human escalation since the global rollout. The training data approach made the difference. Initial SOPs produced 75% accuracy. Once the team restructured training data around the specific questions employees were actually asking, accuracy hit 93%. The Pittsburgh location recently logged 300 interactions in a single week at 100% accuracy. Onboarding e-learning dropped by 15 to 20%. New store teams no longer train on every possible issue. Trainers focus on fundamentals and Apollo handles the long tail, acting as a second trainer on the floor during new store openings. Chatbase's multilingual capabilities meant Apollo could serve locations in Tokyo, Zurich, and Italy from a single English training set, responding in whatever language the question was asked in. Even the skeptics came around. The deployment team, who build Sandbox VR's hardware, were initially anti-AI. The pilot's 3,000 messages and accuracy data changed their minds.
“All it took was the 3,000 messages in our pilot to prove to that team, not only is this helpful, but it is very, very accurate. And once they saw that accuracy rating from our knowledge agent, they were fully bought in.”
More customer stories
How Aplazo doubled merchant acquisition without adding headcount
How Jumia J Force scaled support to 8 African markets with Chatbase
How West Coast Batteries solved complex battery recommendations with Chatbase
Better support starts today
Join thousands of teams using Chatbase to build AI agents that deliver exceptional support at scale — across chat, email, and voice.