Ticket volume cut by 55% - without an extra support team
A case study with Rheinische Metallbau AG, Cologne
Introduction
Markus Weber heads the field service at Rheinische Metallbau AG, an industrial company with 45 employees in Cologne. The problem: The support team was drowning in recurring customer questions - delivery times, product specifications, assembly instructions. At the same time, internal employees were digging through emails and files for hours to find answers to process questions. The Support Deflection Agent answered customer questions directly from the knowledge base, the Knowledge Assistant gave employees immediate answers with sources cited.
The Challenge
Rheinische Metallbau had a two-fold knowledge problem: Externally, customers sent 40-50 questions per day by email and phone - always the same questions about delivery times, technical data and assembly notes. The 3-person support team needed an average of 6 hours per answer. Internally, employees searched daily for policies, process documents and product specifications - scattered across emails, SharePoint folders and private notes. 20 employee-hours per week were spent on internal search battles.
Support Deflection Agent + Knowledge Assistant: Knowledge now, not hours later
Instead of hiring 2 new support employees (€70,000/year), Markus implemented two AI agents: The Support Deflection Agent answered customer questions around the clock directly from the existing knowledge base. The Knowledge Assistant gave employees immediate answers on internal policies, processes and products - always with sources cited.
- Support Deflection Agent: Answers customer questions automatically from the product database, FAQ and past support tickets - 24/7
- Knowledge Assistant: Answers employee questions on policies, processes and products - never with invented answers, always with sources cited
- Ticket volume demonstrably drops 30-60%, as the agent answers 70% of recurring questions directly
- Integration into the existing email infrastructure and internal knowledge database (Confluence/SharePoint)
Setup in 7 days, effect from week 1
The implementation was structured and showed first results surprisingly quickly.
Day 1-3: Knowledge-base analysis
Kick-off call. Analysis of the past support tickets (last 6 months), identification of the top 20 questions. Upload of 200 past support responses and all internal process documents for training.
Day 4-6: Agent setup & training
Configuration of the Support Deflection Agent for customer communication (professional, factual). Training of the Knowledge Assistant on internal documents with source links. Test phase with 10 real customer questions.
Day 7: Go-live
Live activation of both agents. On day one the Support Deflection Agent answered 23 of 41 incoming questions directly - without human intervention.
Faster, fewer tickets, happier customers and employees
In 4 months the support team halved ticket volume and employees found internal information in seconds instead of hours.
55%
Fewer support tickets
2 min
Avg. response time (vs. 6 h)
20 h
Weekly time saved (internal)
92%
Customer satisfaction (vs. 74% before)
The math: AI agents implemented once → 55% fewer tickets, 20 fewer hours of internal search battles per week, customer satisfaction up from 74% to 92%. The 3 support employees could focus on complex edge cases instead of repetitive questions. No new staff needed.
What our customers say
"Our team answered the same questions every day - delivery times, technical data, assembly instructions. Now the agent answers 70% of them automatically, and our people focus on the real problems. Internally, employees find answers in 2 minutes instead of 2 hours. That was a game-changer for us."
Markus Weber
Head of Field Service, Rheinische Metallbau AG, Cologne
