AI Assistant ("the Rhino")
The AI Knowledge Assistant (nicknamed "the Rhino") is a conversational chatbot available to all staff. It can answer questions about tickets, clients, historical work, and common procedures by searching through ticket history using semantic understanding. It also supports organization-level issue analysis and is context-aware of the ticket you're currently viewing.
Accessing the AI Assistant
Floating Widget (Quick Access)
A chat bubble appears in the bottom-right corner of every staff panel page. Click it to open a compact chat window.
- Always available while browsing tickets
- Minimizes back to a bubble when closed
- Maintains conversation context across page navigations
- Context-aware — knows which ticket you're viewing and your name
Full-Page View (Extended Sessions)
For longer conversations or complex research:
- Navigate to Staff Panel → AI Assistant (
scp/ai-assistant.php) - Full-width chat interface with sidebar showing conversation history
- Better for multi-turn research sessions
Using the AI Assistant
Asking Questions
Type a question in natural language. Examples:
| Question Type | Example |
|---|---|
| Current ticket | "What's the issue with this ticket?" |
| Ticket-specific | "What was the resolution for the checkout bug last month?" |
| Client-specific | "Has Acme Corp reported this issue before?" |
| Org analysis | "What are the top issues for Acme Corp?" |
| Procedural | "How do we handle refund requests?" |
| Technical | "What's the API endpoint for product sync?" |
| Historical | "When did we last update the payment gateway for ScentLok?" |
Ticket Context Awareness
When you're viewing a ticket, the AI automatically knows:
- Which ticket you're looking at (ticket number)
- Who you are (your staff name)
This means you can ask questions like:
- "What's this ticket about?"
- "Has this client had similar issues?"
- "Summarize the conversation so far"
Without needing to specify a ticket number — the AI references the one you're currently viewing.
Organization Issue Analysis
Ask the AI to analyze patterns across all tickets for a specific organization:
| Query | What It Does |
|---|---|
| "Top issues for Acme Corp" | Identifies top 20 recurring themes |
| "Common problems at ScentLok" | Categorizes tickets by issue type |
| "What does this org usually have issues with?" | Uses current ticket's org |
| "Top issues for Acme Corp in the last 6 months" | Time-filtered analysis |
The AI analyzes up to 500 tickets and returns:
- Issue themes ranked by frequency
- Brief description of each theme
- Example ticket numbers for reference
- Time period covered
How It Responds
- Your question is converted to a vector embedding
- Pinecone searches for the most relevant ticket content
- Context from matching tickets is assembled
- Gemini generates a response using that context
- Response is streamed to you in real-time (word by word)
Tips for Better Answers
- Be specific — "What caused the 500 error on ScentLok checkout in March?" beats "What was that bug?"
- Include client name — helps narrow the search to relevant tickets
- Mention timeframes — "last week", "in Q1 2026"
- Ask follow-ups — the assistant maintains conversation context
- Use "this ticket" — the AI knows which ticket you're viewing
Conversation Management
Starting a New Conversation
Click New Conversation to start fresh. The AI won't have context from the previous conversation.
Viewing History
The sidebar (full-page view) or dropdown (widget) shows your conversation history:
- Conversations are listed by title (auto-generated or your first message)
- Click any conversation to resume it
- Each conversation maintains its own context
Deleting Conversations
Click the trash icon next to a conversation to delete it. This cannot be undone.
Understanding AI Fields on Tickets
The AI system automatically analyzes tickets and populates these fields:
Urgency Badge
| Level | Color | Meaning |
|---|---|---|
| LOW | 🟢 Green | Routine request, no time pressure |
| MEDIUM | 🟡 Yellow | Standard priority, normal workflow |
| HIGH | 🟠 Orange | Needs attention soon, potential impact |
| CRITICAL | 🔴 Red | Immediate attention required, major impact |
Urgency Percentage
A numeric score (0–100) providing finer granularity:
- 0–25: LOW territory
- 26–50: MEDIUM territory
- 51–75: HIGH territory
- 76–100: CRITICAL territory
Estimated Hours
The AI's estimate of effort required to resolve the ticket. Use as a guideline:
- Compare against budget hours to gauge scope
- Flag to PM if AI estimate significantly exceeds budget
- Actual time may vary — this is an estimate, not a commitment
AI Summary
A condensed overview of the ticket covering:
- What the client is asking for
- Key technical details
- Current status of the issue
- Any blockers or dependencies
AI Timeline
A chronological summary of events:
- When the ticket was created
- Key messages and responses
- Status changes
- Escalation events
- Resolution steps taken
Suggested Title
The AI may suggest a better, more descriptive title for the ticket. If you see a suggestion:
- Review it for accuracy
- Click Apply to update the ticket subject
- Or ignore if the current title is fine
When the AI Gets It Wrong
The AI is helpful but not perfect:
| Situation | Action |
|---|---|
| Urgency seems too low | Override the field manually; PM will see your override |
| Urgency seems too high | Override it; this recalibrates over time |
| Summary is inaccurate | Ignore it; it will update on next thread entry |
| Wrong estimated hours | Override with your own estimate |
| Assistant gives wrong answer | Rephrase your question or check the ticket directly |
AI field overrides are preserved — the system won't overwrite manual changes on the next processing cycle.
Force Re-Embed
If you notice the AI Assistant doesn't know about recent ticket updates:
- Open the ticket
- Find the "Re-embed" option (or use the AI Assistant interface)
- Click Re-embed to force the ticket's content to be re-vectorized
- Wait 1–2 minutes for processing
- Ask your question again
This is useful after major ticket updates that the automatic embedding cron hasn't processed yet.
Privacy & Data
- The AI only searches tickets accessible within the ticket system
- Conversation history is stored per-staff member
- The AI cannot access external systems directly
- All queries and responses are logged for system improvement
- Internal notes (type N) are included in the AI's knowledge base
- Client-facing responses should always be written by humans, not copied from the AI