How Salesforce Built a Scalable AI Puzzle App in Six Weeks

The episode follows Jagan, a Salesforce architect, who led the creation of an online puzzle app built entirely on the Salesforce platform and delivered through a custom Slack bot. What was originally a nine‑month roadmap was compressed into just six weeks, launching on February 8 for a live event that attracted millions of participants. The team accelerated development with Agent Force Vibes for rapid prototyping and used Salesforce’s Scale Testing product to simulate concurrent traffic, ensuring the solution could handle a million‑user load without performance degradation and compliance checks throughout development.

Scalability alone wasn’t enough; the project demanded robust AI governance. Jagan’s team designed guardrails that prevented the bot from revealing puzzle answers and integrated automated toxicity detection to block malicious prompts. Recognizing AI’s limits, they added a human‑in‑the‑loop for final answer validation, blending machine speed with expert oversight. A security‑first mindset guided negative testing and advisory simulations, exposing potential prompt‑injection attacks before launch. This layered approach delivered a trustworthy experience while maintaining the speed required for a high‑stakes, real‑time competition and continuous monitoring to maintain safety.

The conversation offers practical takeaways for Salesforce admins aiming to build large‑scale, AI‑enhanced solutions. Leveraging Slack Canvas, lists, and custom workflow automations kept a 14‑member team aligned and allowed weekly status summaries despite the six‑week deadline. Admins are encouraged to start with a security mindset, prototype with Agent Force, and employ Scale Testing early to surface bottlenecks. By combining low‑code tools, rigorous testing, and human oversight, modern admins can move beyond simple automations to architect resilient, responsible systems that scale to millions of users and measurable ROI for the enterprise.