AI IN PRACTICE / SERVICE & SELF-SERVICE

Service & self-service

Empower customers. Prepare field visits.

Illustrative process scenarios — not customer references. Feasibility, benefits and approvals need to be established for the specific organisation.

09

ILLUSTRATIVE SCENARIO

Guide customers through approved self-service tasks

A customer portal offers help matched to the installed machine, collects missing information and hands the complete context to service when needed.

UNDERSTAND THE AI

What happens behind the scenes.

TECHNICAL INSIGHT
POSSIBLE SOLUTION DESIGN

Conversational assistant · Configuration-aware RAG · Guided workflow

What does the AI actually do?

AI understands the operator’s question, asks for relevant missing details and explains approved guidance. It retains the context needed for a complete service escalation.

How the solution works

  1. 01
    Enforce context boundaries

    After sign-in, the portal and asset register supply permissions, serial number and software revision. Retrieval can access only records valid for that person and configuration.

  2. 02
    Choose an approved step

    An LLM interprets the question and retrieves relevant passages. A defined workflow controls allowed steps and stop conditions; missing sources trigger clarification or handover.

  3. 03
    Capture the outcome

    Responses, observed messages and completed operator checks are recorded in structured form and can become a service-ticket draft.

What value can it deliver?

Faster resolution of permitted standard questions for customers; more complete tickets and less repeated questioning for service. Repeat-contact rate helps establish whether reported resolutions last.

A concrete example

Illustration: two manual variants describe the same message differently. The asset register restricts retrieval to the installed software revision, and only the applicable approved steps appear.

Technology, setup & operation

The design combines a pretrained LLM, retrieval index and dialogue-state workflow. Backend software enforces permissions and allowed actions. Ongoing quality checks cover revisions, unsupported answers and escalation behaviour.

Illustrative solution design. The final architecture depends on data, systems and pilot evaluation.

Starting situation

A machine operator wants to resolve a recurring operating message. The general manual covers several variants. The portal should provide the exact approved steps for that machine rather than inventing repair instructions.

How the workflow could look

  1. Confirm machine and access

    After sign-in, the serial number, software version and permitted operator tasks are matched. Unidentified machines are referred to service.

  2. Provide guided support

    The assistant explains the relevant approved steps and collects results in a structured form. Sources and document revisions remain visible.

  3. Transfer context

    If the issue persists or a defined stop condition occurs, the portal prepares a ticket with the message, machine version and checks already completed.

Illustrative work product

Machine-specific guidance or a complete service handover, so the customer does not need to repeat the entire explanation.

10

ILLUSTRATIVE SCENARIO

Prepare field visits with the right parts and machine context

Before travel, a visit pack combines installed configuration, case history and confirmed spare parts. After the visit, it supports a reviewable service report.

UNDERSTAND THE AI

What happens behind the scenes.

TECHNICAL INSIGHT
POSSIBLE SOLUTION DESIGN

RAG knowledge assistant · Structured extraction · Report drafting

What does the AI actually do?

AI condenses long asset and case histories for a specific visit, retrieves similar incident descriptions and drafts a report from the engineer’s actual observations.

How the solution works

  1. 01
    Link asset and history

    Serial number and configuration constrain retrieval. Semantic search finds differently worded cases; modifications and part supersessions come from authoritative systems.

  2. 02
    Prepare the visit pack

    An LLM summarises the relevant evidence. Approved part relationships and ERP queries check compatibility and stock; uncertain suggestions are flagged.

  3. 03
    Turn entries into a report

    Notes or approved voice transcripts populate fields such as finding, action and parts used. Missing measurements stay empty and prompt a request for completion.

What value can it deliver?

Less preparation and documentation effort, with a lower risk of arriving with unsuitable parts or incomplete records. Better preparation can reduce avoidable repeat visits.

A concrete example

Illustration: an earlier repair mentions an old part number. An approved supersession identifies its replacement; service checks this against recorded modifications before allocating it.

Technology, setup & operation

Pretrained language and search models connect to service and ERP interfaces. Structured domain rules govern part approval. Report faithfulness is assessed against the engineer’s entries; a similar historical case is not a confirmed diagnosis.

Illustrative solution design. The final architecture depends on data, systems and pilot evaluation.

Starting situation

A field engineer visits an older custom machine. Modifications, previous repairs and part substitutions sit in different systems. Missing tools or an incompatible spare-part variant may only become apparent on site.

How the workflow could look

  1. Consolidate context

    The assistant brings machine configuration, modifications and relevant incidents into an evidenced visit overview. Conflicts with the documented installed state remain open.

  2. Check preparation

    Approved procedures and parts lists inform proposals for tools, qualifications and spares. Service staff confirm compatibility and availability.

  3. Document on site

    The engineer records observations and actions in a structured form. The assistant drafts a report and flags missing values before expert approval.

Illustrative work product

An approved visit pack with sources and parts, plus a report based on actual entries. A version-bound offline pack can be provided.

FROM SCENARIO TO A PRACTICAL NEXT STEP

Which process deserves
a closer look?

Together, we can clarify the current process, available evidence and a useful pilot. My contribution: structure, priorities and alignment between business teams, IT and implementation partners.

Discuss a use case