Why this system exists
Most support tools only store tickets. This one actually investigates them, talks the request through with the people involved, and carries it all the way to a reviewed, deployed fix. When a user asks “Is the Booking Source field available? Can we get it added to some reports?” the system does not just create a ticket. It:
- Understands the request and checks the real database for the field
- Asks a clarifying question, then turns the answer into a ticket
- Lets the assigned developer investigate and confirm the change with AI
- Shows the proposed code changes to that developer only
- Applies the change with Claude, on the developer's instruction
- Routes the commit through QA regression tests and manager approval
- Leaves the final "it's resolved" call to the user who asked for it
System Explainer Video
Auto-playing cinematic walkthrough of the complete flow.
Smart AI Support
Multi-Agent Ticketing System
OpenAI • Claude • Human-in-the-Loop
User
Which reports would you like it added to?
User
Sales Report and Booking Summary Report, please.
We've created your ticket and assigned it — we'll be in touch once it's ready.
This ticket needs Booking Source added to Sales Report and Booking Summary Report — can I go ahead?
Yes — Booking Source exists in the domain model and is unused in both report DTOs today. Safe to add.
Add BookingSource to SalesReportDto and BookingSummaryReportDto, and map it through both report query services.
SalesReportDto.csBookingSummaryReportDto.csReportQueryService.csdotnet test…Human stays in control
Agents investigate. Claude drafts and applies.
Developers decide. QA verifies. Managers approve. Users confirm.
Architecture at a Glance
OpenAI / Azure OpenAI
Understanding, reasoning & root-cause synthesis
Claude (Anthropic)
Code-fix drafting specialist (diffs only)
OpenAI
Multi-Agent
OpenAI
Claude
Developer + Admin
Live Interactive Demo
Click through the exact story from the video above yourself — every reply below is scripted to match it exactly, step by step, so you can pace it at your own speed.
Core Design Principles
Read-only Investigation
Agents only read code, database and logs. Nothing is written during analysis.
Human-in-the-Loop
Claude only drafts. Real code changes and deployment require human approval.
Real Regression Tests
Every fix goes through actual dotnet test.
Progressive Intelligence
Works without LLM (heuristic fallback). Becomes much smarter when AI is available.
Who Provides Accurate Code?
| Need | Responsible | AI Involved |
|---|---|---|
| Understand the ticket | TicketIntakeService | OpenAI / Azure OpenAI |
| Identify bugs / gaps | Multi-Agent Investigation | OpenAI / Azure OpenAI |
| Explain root cause | RootCauseEngine | OpenAI / Azure OpenAI |
| Propose accurate code fix | ClaudeFixDraftingService | Claude |
| Approve the changes and move the updated file to dev repository | Human Developer | Assisted by Claude |
| Test & Deploy | System + Admin | None |