Miljenko Vukovic

For years, support and maintenance was the job nobody wanted — and somehow it always ended up on the desk of the engineer who was best at solving problems. In one B2B software business, around 45% of support requests were outside the maintenance contract, ranging from data investigations and system issues to customers asking for help with their own work. Agentic AI changed the equation. In this talk, we will share what we learned from automating support and maintenance in production: how to classify the different kinds of support requests, which ones an AI agent can safely handle on its own, and which ones should remain with humans. We will look at practical patterns for customer-specific agents, coding agents, data investigation, and first-line support — as well as the guardrails that prevent an autonomous system from confidently doing the wrong thing. The key lesson is surprisingly simple: the first thing support automation should eliminate is not the difficult engineering work, but the unbillable tide of repetitive, ambiguous and out-of-scope requests that consumes your best people. Find out more at: https://vis-solutions.eu/en/blog/the-support-agent-that-took-my-worst-job-away