"A technician with limited product knowledge found the right solution through the Opero AI agent and resolved the issue without interrupting a colleague who was out with customers."
The starting point
Eighteen years of service knowledge — none of it searchable
Nize Equipment sells and services large-format print and finishing machinery across the Nordics. Their service organisation is small by design: lean, senior, and constantly on the road. When a new technician joins, the ramp-up curve is steep — most of the institutional knowledge lives in PDFs scattered across a network drive, in OEM service bulletins emailed from Europe, and in the heads of two or three senior engineers.
In 2024 the service-and-IT lead, Mette Meyer Thuesen, mapped the cost of that fragmentation. Roughly one in three field calls became a phone call back to a senior engineer. Onboarding a new technician took six months before they were billable on complex jobs. Customer downtime was the visible KPI; internal escalations were the hidden one.
The approach
Index every manual. Cite every answer. Put it in front of the engineer.
We started with the corpus: 1,400 service documents across three OEM product families, two languages, and roughly twelve years of revision history. The first deployment was an internal-only agent — chat in the browser, citations back to the source PDF, and a feedback widget on every answer.
Within four weeks the agent was answering questions a technician would otherwise have escalated. The interesting work was downstream: tagging revisions so the agent never cites a superseded bulletin, surfacing model-specific schematics, and adding a voice mode so engineers on the floor could ask hands-free.
The result
A 60% drop in escalations, and a senior team that gets to think again
Ninety days in, L2 escalations had dropped by 60%. New hires were reaching first-call independence in roughly six weeks instead of six months. The senior engineers — the ones who used to be the bottleneck — got their afternoons back.
Phase two added the customer portal: end-customers ask the same agent before opening a ticket. Phase three, currently in pilot, adds parts ordering: when the agent identifies the failing component, it pre-fills an order against the right SKU. Mette’s team is the customer of the agent. Their customers are the customer of the portal. The senior engineers are the customer of their own time back.