Title

Automated Ticket Classification & SLA Escalation for a 4-Location Barbershop Chain — Zero Missed Priority Alerts

Context

BRO Barbershop is a chain with four locations. At each branch, administrators combine handling online inquiries (via Telegram, website chat, and phone calls) with in-person customer service. The chain receives approximately 400 inquiries per month, 65% of which relate to booking appointments, checking prices, or rescheduling and cancellations. Prior to implementation, ticket classification and prioritization were handled manually—based on estimation—without a formal SLA or escalation mechanism.

Industry: Beauty & Personal Care Services (Barbershop)

Problem

• Administrators couldn't keep up with timely responses; approximately 30 bookings were lost last month due to slow replies.
• There was no unified criterion for prioritizing tickets; everything was handled informally, on a first-come, first-served basis.
• NPS dropped from 56 to 38 over six months; complaints got lost among routine inquiries and failed to receive the necessary priority attention.
• The owner considered hiring a dedicated administrator, even though the software budget was limited ($100/month).

Solution

Two interconnected workflows have been built in n8n:

1. AI Ticket Classifier — an agent classifies incoming inquiries into five categories (new record, record modification, pricing, complaint, other), assigns a priority level (P1–P4) using chain-of-thought reasoning (urgency → business impact → customer context), and determines the default owner.
2. SLA Watcher & Escalation — monitors open tickets every minute, calculates deadline breaches, and triggers automatic escalation across three levels (Owner → Team Lead → Director) via Slack and email.

Process

1. Problem & Data Source Analysis

I started by mapping how tickets actually flowed in across all four locations — Telegram, website chat, phone calls — and found that administrators were prioritizing purely by gut feel and order of arrival, with no shared definition of what counted as urgent. This explained both the lost bookings and the dropping NPS: complaints were getting buried under routine pricing questions.

2. Designing the Classification & Priority Logic (Prompt Design)

I defined five ticket categories (new booking, booking modification, pricing, complaint, other) and a P1–P4 priority scale. Instead of a flat classification prompt, I built in chain-of-thought reasoning — urgency → business impact → customer context — so the AI weighs a case the way an experienced manager would, not just by keyword matching. I also mapped each category to a default owner, so tickets don't sit unassigned.

3. Integrating Input Channels with AI (API Integration)