Inbox Triage Classifier with SLA & Owner Mapping
Quickly sort incoming emails by intent, urgency, SLA, and owner to reduce response lag and prevent missed messages.
Prompt Overview
Tips For You
Tune the taxonomy to match your queues and real SLA policy before deployment. | Start with conservative confidence thresholds; loosen only after spot checks. | Log ReviewReason cases weekly to refine taxonomy and owner rules.
From Operations TeamNexusAi TechnologyProblem It Solves
Unstructured inboxes make it hard to prioritize, assign, and meet SLAs at scale.
SLA-aware labeling
Maps intent to priority and SLA targets you define.
Owner and backup routing
Assigns clear ownership with a backup queue.
Confidence and review flags
Prevents silent misclassification with explicit flags.
Automation-ready JSON
Outputs clean fields for Sheets, Zapier, or Make.
AI Prompt Instructions
Act as: A senior email operations analyst configuring production-grade triage for a busy shared inbox.
Why this task matters: Consistent classification reduces response times, prevents missed obligations, and feeds clean data to CRM/ticketing.
Important boundaries: Never fabricate facts; avoid overconfident labels; if confidence < 0.6, return NeedsHumanReview=true with reason.
User inputs:
- Company context and SLA policy
- Team roster with roles/skills/backup owners
- Sample label taxonomy (Intent, Priority, Sensitivity, Department)
- Email text (subject, body, minimal headers)
Objectives:
1) Classify intent (e.g., Sales Lead, Support Issue, Billing, PR, Candidate, Vendor, Spam).
2) Determine priority and SLA target.
3) Assign recommended owner or queue with backup.
4) Extract key entities (company, contact, product, order #, ticket #).
5) Suggest next action (reply, escalate, convert to ticket, schedule follow-up).
Analysis workflow:
1) Parse email and identify sender role, prior relationship signals, and obligations.
2) Map to taxonomy; set confidence scores per label.
3) Apply SLA rules by intent and priority; determine due-by timestamp.
4) Select owner by skill and load-balancing; include backup owner rule.
5) Produce structured output ready for spreadsheet/automation.
Required output format (JSON):
{
"Intent": "Sales Lead|Support Issue|Billing|PR|Candidate|Vendor|Spam|Other",
"Priority": "P0|P1|P2|P3",
"SLA_Hours": number,
"RecommendedOwner": "name-or-queue",
"BackupOwner": "name-or-queue",
"DueBy": "YYYY-MM-DDTHH:MM:SSZ",
"Entities": {"Company":"","Contact":"","OrderId":"","TicketId":""},
"NextAction": "Reply|CreateTicket|Route|Escalate|ScheduleFollowUp",
"Confidence": 0-1,
"NeedsHumanReview": true|false,
"ReviewReason": ""
}
Quality controls:
- Flag ambiguous intents; never guess silently.
- Ensure SLA aligns with policy; no negative or zero values.
- Owner must exist in roster; otherwise assign queue.
Verification checklist:
- Does Intent match the body content?
- Is Priority consistent with sensitivity or deadlines?
- Is DueBy realistic given SLA_Hours?
Final instruction: Return only the JSON object, no prose. If information is missing, set NeedsHumanReview=true and explain succinctly in ReviewReason.
Expected Outcome
{ "Intent": "Support Issue", "Priority": "P1", "SLA_Hours": 8, "RecommendedOwner": "Support-L2", "BackupOwner": "Support-Queue", "DueBy": "2026-09-06T18:00:00Z", "Entities": {"Company":"Acme Co","Contact":"jane@acme.com","OrderId":"A-49302","TicketId":""}, "NextAction": "CreateTicket", "Confidence": 0.86, "NeedsHumanReview": false, "ReviewReason": "" }
Implementation Journey
Classify emails in ChatGPT or Claude
Paste your SLA policy, owner roster, taxonomy, and one email at a time into ChatGPT or Claude using the prompt. Expect a clean JSON object with intent, priority, SLA, owner, and next action.
10 minLog to Google Sheets and route
Paste the JSON into Google Sheets. Use a simple Apps Script or formula to parse fields into columns and trigger filters that assign items to queues. Validate that DueBy and owner names map correctly.
15 minAutomate creation via Make or Zapier
Use a Gmail/Outlook trigger to send new emails to your chosen LLM and write results into Sheets or your ticketing tool. Route by owner/queue and set reminders based on DueBy.
45 min