Ops Intake Router Prompt (Unstructured Requests → Triage & Entity Extraction → Assigned Ticket & SLA Plan)
Turn emails, chats, and form posts into structured tickets with priority, owner, due date, and SLA—so ops teams stop firefighting and start executing.
Prompt Overview
Featured AI Partner
Tips For You
Start with 20–30 varied samples to train robust categories and entity patterns. | Keep categories under 30 and priorities under 5 for fast operator decisions. | Use short, distinct tags to enable later automation triggers. | Add business hours and holiday calendars to improve due_at accuracy.
From Operations TeamNexusAi TechnologyProblem It Solves
Manual triage wastes time, loses context, and creates inconsistent SLAs. This prompt standardizes intake and assignment with explicit rules.
Entity extraction for routing
Pulls customer, order, system, and region from messy text.
SLA-aware deadlines
Sets due dates aligned to priority, calendars, and business hours.
Ownership suggestions
Proposes owners using skills and load when available.
Risk flags & clarifications
Prompts for missing data and highlights potential compliance issues.
AI Prompt Instructions
Act as: Senior Operations Intake Manager and Systems Designer.
Why this task matters: Unstructured inbound requests (email, chat, forms) stall execution. Fast, consistent triage with clear ownership, SLAs, and sequencing reduces cycle time and errors.
Important boundaries:
- Never invent facts; extract only from provided text and well-defined routing rules.
- If mandatory fields are missing, request clarification using a concise template.
- Respect PII handling rules and mask sensitive data when outputting.
User inputs:
- Inbox sample (10–50 messages) or daily dump of requests.
- Routing rules: categories, priorities, SLA matrix, business hours, holidays.
- Team roster with skills and capacity (optional).
Objectives:
1) Normalize each request into a structured record.
2) Classify category, priority, and intent; extract entities (customer, product, region, system).
3) Propose owner and due date aligning with SLA rules and capacity.
4) Highlight blockers and missing info.
Analysis workflow:
1) Parse messages; detect language; summarize in 1–2 sentences.
2) Classify using rule+semantic cues; map to category and subcategory.
3) Determine priority via impact×urgency; map to SLA target (hours/days).
4) Extract entities (customer_id, order_id, system, channel, severity, attachments).
5) Propose owner by skill tag and load (if capacity provided); else route to default queue.
6) Detect compliance/PII; mask as needed.
7) Build clarification prompts for low-confidence fields.
Required output format:
- JSON array of records with: id, summary, category, subcategory, priority, confidence, entities{customer_id,order_id,system,region}, sla_target, due_at, proposed_owner, required_clarifications[], tags[]
- Include a separate routing_summary: totals by category, priority, and owner.
Quality controls:
- Confidence <0.7 must include at least one clarification.
- Validate due_at against business hours and holidays.
- Flag duplicates and link related tickets.
Verification checklist:
- Are SLA targets consistent with priority?
- Are PII fields masked where required?
- Do owners map to known team members?
Final instruction: Produce only the JSON outputs. No prose beyond a 3–5 line operator note with key risks and next actions.
Expected Outcome
{ "routing_summary": {"total": 24, "by_priority": {"P1": 3, "P2": 8, "P3": 13}}, "tickets": [ {"id": "REQ-2026-0712-01", "summary": "Invoice sync failing for ACME EU", "category": "Billing", "subcategory": "SyncError", "priority": "P1", "confidence": 0.86, "entities": {"customer_id": "ACME-221", "order_id": null, "system": "ERP-1", "region": "EU"}, "sla_target": "4h", "due_at": "2026-07-03T16:00:00Z", "proposed_owner": "ops-billing-oncall", "required_clarifications": [], "tags": ["incident","finance"]} ] }
Implementation Journey
Generate the triage schema in ChatGPT or Claude
Paste 20–50 real inbound messages plus your routing rules (categories, priority logic, SLA matrix) into ChatGPT or Claude with this prompt. Ask for a JSON schema and 5 sample classified records to validate entity extraction and SLA mapping.
15-20 minDeploy the Intake Router in OpenWork
In OpenWork, create an agent called Intake Router. Configure an input node for raw messages and a processing node that applies the JSON schema fields returned by the model. Map outputs to fields: summary, category, priority, entities, sla_target, due_at, proposed_owner.
20-30 minPublish and connect channels
Publish the OpenWork agent and connect email or chat webhooks. Test with 5 live messages, verify due_at within business hours, and confirm owner mapping. Capture metrics on classification confidence and time-to-first-response.
20-30 min
