SupportAI

Overview

SupportAI is a production-minded AI customer support SaaS demo built as a full product, not just a chatbot widget. It combines a public-facing marketing website with an authenticated internal dashboard for support teams. The core idea is simple: let AI handle common questions, keep it inside clear guardrails, and send low-confidence or higher-risk cases to a human team.

What this demo shows

• A complete SaaS product experience (marketing site plus protected app area)
• Role-based support operations (different access for admins, support managers, and agents)
• AI support workflows with human handoff when confidence is low or the topic needs escalation
• A maintainable full-stack structure (UI, validation, services, and data access kept separate)
• Security-minded product design (protected routes, safe input handling, rate limiting, and controlled error responses)

What teams can do

• Browse a public SaaS website with landing, product, solutions, security, pricing, privacy, and terms pages
• Test an AI support sandbox across five business scenarios: SaaS, construction, dental, e-commerce, and consulting
• Sign in with credentials and access protected support routes based on role
• Review conversations with search, filters, sorting, pagination, transcript views, internal notes, feedback tracking, assignment, resolution, and escalation actions
• Manage an escalation inbox with urgency indicators, priority levels, overdue states, assignment controls, and resolve actions
• Create, edit, duplicate, archive, and test knowledge base articles in a simulated answer flow
• Customize the support widget with live preview, theme controls, prompt chips, escalation toggle, business-hours messaging, and generated embed code
• Review analytics for conversation volume, resolved vs escalated cases, feedback trends, unanswered topics, and widget engagement
• Manage settings for workspace behavior, notifications, permissions, and future integrations

Business-focused features

• AI guardrails help keep automated support useful without pretending it can safely answer everything
• Confidence scoring and escalation rules support human oversight instead of replacing the team blindly
• The dashboard structure is ready for real support operations, not only a visual concept
• The product is organized so mock data can later be replaced by real providers, APIs, and databases with less frontend rework
• The mix of marketing pages and internal tools shows how a SaaS product can be presented and operated in one consistent system

Demo notes

• Authentication and the Prisma support domain model are real parts of the project
• The AI assistant is currently simulated and not yet connected to a live OpenAI provider
• Most operational data currently runs through a mock repository so the demo stays deterministic and easy to present
• That mock setup is intentional for portfolio use, because it shows product architecture clearly without pretending unfinished integrations are live