App Home Page

ActSafe：AI 驱动的安全事件管理平台

Inspiration

Small businesses manage safety incidents the same way for decades: someone notices a problem, writes it on paper, calls a manager, and hopes it gets followed up. We asked: what if AI could automate the analysis and response planning, but humans stay in control of every execution decision? That's "Authorized to Act."

What it does

ActSafe is an AI-powered safety incident management platform:

Incident Dashboard – real-time stats + AI summaries
Record Incident – 30-second reporting, AI generates action plan
Action Center – human approve/reject every AI-suggested action
Audit Log – full traceability
Settings – Auth0 + Discord integration

How we built it

Frontend: Next.js 16 + TypeScript + Tailwind CSS
Authorization: Auth0 Token Vault (authorization layer)
AI Engine: OpenAI GPT‑4 (action plan generation)
Notifications: Discord Webhook
Key design:
- Dual AI engine (OpenAI + local rule fallback)
- Auth0 as execution gateway (not just login)
- Card-based human-in-the-loop UI

Challenges we ran into

Auth0 depth – using Token Vault to authorize each action execution was architecturally complex
UI density – stats + lists + AI analysis in one screen, iterated 3 design versions
AI reliability – LLM occasionally generates unrealistic plans; fallback rule engine ensures always-on

Accomplishments that we're proud of

Complete incident-to-audit workflow built solo in 3 days
Auth0 goes beyond authentication – it’s the execution authorization layer, perfectly matching "Authorized to Act"
Reusable AI + human-in-the-loop pattern for any escalation workflow

What we learned

AI shouldn’t have direct execution power – AI-analyzes, human-decides, system-executes
Auth0 Token Vault can be an authorization gateway for every API call
Fallback design is essential for AI-powered systems
Small teams deserve enterprise-grade tools at low cost