Instead of a simple assistant, build a Cross-Platform Compliance & Action Layer. Sentinel lives in the browser (Track 1) but orchestrates deep backend workflows (Track 2). It monitors what an employee is doing in real-time (e.g., writing a sensitive email or viewing a contract) and proactively triggers "Guardrail Agents" to ensure data safety while automating the next five steps of the workflow.🛠️ Detailed Breakdown1. Technological Implementation (The "Quality" Proof)The Stack: Use Airia’s Nested Agent Architecture.The "Panadol" Feature: Implement a PII (Personally Identifiable Information) Redaction Agent that sits between the User and the LLM.Code Quality: Use a modular approach.Agent A (The Scout): A browser extension that scrapes the current DOM context.Agent B (The Judge): Uses Airia's Human-in-the-Loop (HITL) feature to ask for permission before moving data between systems (e.g., from Outlook to a CRM).Formula for Accuracy: Use a RAG (Retrieval-Augmented Generation) pipeline within Airia to ensure the agent only uses company-approved documentation:$$\text{Reliability} = \frac{\text{Verified Data Points}}{\text{Total Agent Actions}} \times 100$$2. Design & User ExperienceThe "Frictionless" UI: Do not build a new dashboard. Build a Sidebar Extension that "ghosts" over Salesforce, LinkedIn, or Outlook.Frontend/Backend Balance: The frontend handles real-time "Action Suggestions" (UI/UX), while the Airia backend handles the heavy lifting of multi-system API calls.3. Potential Impact (The "Investor" Pain Point)Problem: $75\%$ of enterprise AI pilots fail due to "security anxiety" and "tool sprawl."Solution: Your project acts as the Universal Interface. It reduces the need for 10 different Slack bots by centralizing them into one Airia-governed agent.Market Scale: It addresses the $200B+ Enterprise Governance market.
Log in or sign up for Devpost to join the conversation.