AI Agent
Arabic and English answers grounded in approved company knowledge, with uncertainty routed to a person.
A full-stack customer experience platform connecting an AI agent, company knowledge, shared conversations and human support in one system.
Open live product ↗Repeated questions create unnecessary workload when information is scattered across documents, web pages and individual conversations. A disconnected handoff also forces customers to repeat context.
I built TAWADUD around one connected workflow: the AI answers from approved knowledge, conversations stay visible to the team, and a person can take over without losing history.
Arabic and English answers grounded in approved company knowledge, with uncertainty routed to a person.
A shared view of conversations, assignment, follow-up and replies from human teammates.
Import and index website pages, documents and written answers for retrieval during conversations.
Move a conversation from AI to a person while keeping its history in the same workspace.
The application combines Next.js and React interfaces with TypeScript product logic, Supabase/PostgreSQL data, workspace access controls, OpenAI-powered answers and Vercel deployment.
I implemented the embeddable website widget, knowledge ingestion flows, conversation assignment, analytics, workspace roles, multi-factor authentication and account security controls.
I also tested customer and operator paths, reviewed failure states and iterated on behavior until the product worked as one coherent system rather than a set of disconnected screens.