How Beast Industries Used Agentforce to Power a 1M-Player Super Bowl Puzzle Hunt
Beast Industries is the creator-led conglomerate behind MrBeast, one of the world’s most-followed YouTube channels with hundreds of millions of subscribers. For Super Bowl 2026, Beast Industries partnered with Salesforce to build a first-of-its-kind AI-powered puzzle hunt that ran for a month, supporting over 1 million active players and 1.5 million simultaneous users at peak. An agentic Slackbot — built on Salesforce’s Trust Layer — guided players through clue discovery without revealing answers, blocking 64,000 prompt injection attacks in the process.
Impact
1 million+
Players who relied on Slackbot’s guidance
1.5 million
Simultaneous users supported at peak
99.997%
Prompt injection attacks blocked automatically
64,000
Total prompt attacks blocked
6 weeks
Platform build time
Challenge
Beast Industries needed to build a secure, real-time AI-powered puzzle hunt platform in six weeks capable of supporting 10 million registrations and 1.5 million simultaneous users, while preventing players from using adversarial prompts to extract the winning answer.
Solution
Salesforce built an agentic Slackbot on Agentforce with Trust Layer security — providing zero data retention, prompt injection defense, and a zero-leakage architecture — that guided over 1 million players through a month-long Super Bowl puzzle hunt without revealing answers.
Tools & Technologies
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.
Full Story
Beast Industries runs high-stakes, ambitious entertainment experiences at global scale. When the company partnered with Salesforce for a Super Bowl 2026 commercial, the brief went far beyond a 30-second spot: the goal was to launch a live puzzle hunt tied to a $1 million prize, with the ad as the starting gun. The resulting platform needed to support tens of millions of potential registrations, millions of concurrent users, and a volume of AI token requests unlike anything Salesforce had deployed in a consumer context.
The core challenge was building at Super Bowl scale in six weeks. The platform had to support up to 10 million registrations, sustain 1.5 million simultaneous users, and process a projected 1 trillion token requests in its first 72 hours. Security was equally critical: with millions of people trying to extract the winning answer through creative prompt manipulation, the system needed to resist adversarial inputs without compromising the game experience.
Salesforce built an agentic Slackbot powered by Agentforce and protected by the Trust Layer. The Slackbot acted as an intelligent puzzle partner: players could share theories, ask questions, and test hypotheses, but the agent was constrained to guide rather than reveal. The Trust Layer provided zero data retention, prompt injection defense, and a zero-leakage architecture. Of the adversarial attempts, 99.997% were blocked automatically, with 64,000 specific prompt attacks flagged and stopped.
The puzzle ran for a month. Over 1 million players relied on Slackbot’s guidance during that time, testing theories and refining their approach to the challenge. At the end, a player slacked the winning message directly to MrBeast — closing a loop that began with a Super Bowl commercial and ended in a collaborative, AI-mediated hunt across a massive global audience.
The case establishes that consumer-facing AI agents can operate at live-event scale with security constraints that match the adversarial pressure of a public audience. For entertainment companies considering interactive AI experiences, Beast Industries’ Super Bowl activation represents a proof point for what’s achievable when platform infrastructure, trust architecture, and creative ambition align.