KI-Assistent
Support AI assistant
A support team that answers in minutes and onboards in days.
- faster troubleshooting
- 80%
- to onboard a new hire
- 30 → 3 days
- tickets resolved
- 500+
- Kunde
- CarOnSale
- Rolle
- AI Automation & Engineering Support
- Jahr
- 2026
Diese Fallstudie gibt es derzeit nur auf Englisch.
CarOnSale had no dedicated engineering support, and every technical question landed on developers. I built the department from scratch, documented what people kept asking, and gave the team an AI assistant that knows the product. Troubleshooting got up to 80% faster, and new hires were productive in 3 days instead of 30.
- 01A support question comes in
- 02The assistant searches the documentation
- 03It points to the likely cause
- 04 · braucht SieThe support agent decides and answers
Every question landed on a developer
There was no team between product and engineering. When something broke for a customer, the question travelled until it reached a developer, who stopped what they were doing to answer it. The same questions came back again and again, because the answers lived in people’s heads.
New support hires had it hardest. It took about a month before they could solve problems on their own.
A department, not a ticket queue
I started by building the engineering support function itself: clear workflows for how issues move between Product and Engineering, and who owns what. I resolved more than 500 tickets along the way, which showed me where the real gaps were.
Then I streamlined the technical documentation, so answers no longer depended on who happened to be online.
An assistant that knows the product
With the documentation in place, I built an AI assistant for the whole technical support team. It answers from the company’s own knowledge, so a support agent can find the cause of a problem without waiting for a developer.
I also automated the upkeep of the product ownership database with n8n, so the question “who owns this?” always had a current answer.
From a month to three days
Troubleshooting became up to 80% faster. New team members were productive after 3 days instead of 30.
Under the hood
- Langdock
- n8n
- Jira