Use Case · Service · 2026

How service cases become available knowledge.

A pump and apparatus manufacturer with its own service · 8 technicians, ~2,000 machines in the field · POC in one week.

Cut downtime, keep the experience, even when the senior technician is not available.

Baseline

Service grown over years. Knowledge in heads and reports.

Pumps, seals, sensors, valves – across many industries and media. The diagnostic experience sits in reports and in the heads of individual technicians.

Service technicians8team in the field

Machines in the field~2,000installed base

Visits / year~600faults & maintenance

Documented cases45symptom · cause · action

Representative profile. Aggregate figures are model assumptions; the service cases and the example below come from real, anonymized cases.

Problem

Three ways today. None cuts the downtime.

When a machine fails, whoever is on site decides, often without knowing that the same fault is already documented.

Way 1Improvise on site

first guess instead of known cause
wrong spare part installed
downtime gets longer

Way 2Search the manuals

knowledge scattered across reports
patterns not recognizable
time while the machine stands

Way 3Call the senior

senior becomes the help desk
not always reachable
lost with turnover

When the first guess is followed instead of the known cause, every misdiagnosis costs a shift plus extended downtime at the customer, often with an SLA behind it.

Architecture

How the agent is built.

Four modules – from reading service reports to the searchable fault pattern. Not a new tool, but structured processing of what is already there.

Input · existing data sources

Service reportsMaintenance manualsSpare-parts listsFault codes

↓

Module 01Reader

Reads and extracts from service reports, manuals, spare-parts lists and .msg. Understands free text.

I/O · raw data → structured text

↓

Module 02Classifier

Recognizes fault attributes by the jointly defined criteria – component, symptom, medium, material, machine.

I/O · text → fields with values

↓

Module 03Dossier builder

Aggregates per fault pattern: symptom, first guess, actual cause, action, spare part, lesson learned.

I/O · fields → complete dossier

↓

Module 04Filter index

Makes all cases searchable by symptom, part, medium and machine. The technician finds the pattern in seconds.

I/O · dossiers → search index

Output · what service uses

Spot patterns instantlyRight spare part first timeCut downtime

Approach

POC setup in one week.

This is what the proof looks like. The POC always runs on a small, bounded data package, for example one device or fault-pattern group. The productive implementation builds on it step by step.

Day 101Data package

Hand over a small, bounded package, for example one device or fault-pattern group. IT effort: ½ day.

Day 202Criteria workshop

Symptoms, components, media and lesson-learned fields defined together. 2-hour workshop.

Day 3–403Agent setup

Reader · classifier · dossier builder · filter index configured. First case dossiers generated.

Day 504Test & handover

Validation on 5 real faults. Walkthrough with the team. Handover to the technicians.

POC effort on the customer side: ½ day. At the end of the week you see what is possible. Whether and to what extent you go productive, you decide afterwards.

Output

What the technician sees in the end.

Ask like in a chat, answers with a source. The agent recognizes the pattern from the service cases and names the actual cause, instead of guessing.

Getze Knowledge Agent·Service Knowledge Agent

v1.0

SERVICE KNOWLEDGE AGENT

Service cases: spot patterns, cut downtime.

POC · small data package

YOU

Service technician

The mechanical seal on our progressing cavity pump keeps breaking after the weekend. What could be the cause?

Service Knowledge Agent · asks instead of guessing

Before I guess, a quick question: what medium does the pump handle?

YOU

Service technician

Glucose syrup.

Service Knowledge Agent

Then the cause is not material fatigue, but crystallization. Over the weekend the glucose crystallizes in the idle pump, and on Monday start-up the seal cracks at the crystal deposits.

Recommended action:
1. Clean the pump after every shift, no residual medium over the weekend.
2. Switch the seal to a hard-material pairing, SSiC/SSiC.

Source: Service case 04 · confectionery industry

Ask a question

Impact

Downtime cut. First-time fix up. Knowledge kept.

Three measurable effects once the service knowledge is searchable across all cases.

Diagnosispattern instantlyknown cause instead of fresh troubleshooting on site.

Spare partright first timeright variant instead of off-brand, fewer follow-up visits.

Knowledgeindependentexperience from 45 cases available, even without the senior.

Use case as PDF

Download the full use case.

Including architecture, the full case example and impact. You get the PDF right after submitting, plus a short email from me.

Next step

Start with a real service case.

We test the value on a real fault pattern from your service, within one week, with a small data package. If the POC doesn't convince you, it costs nothing.

Book a 30-minute call →