In-house Manufacturing EMEA Company newsroom

KION Group

Document Processing

KION's service technicians write maintenance reports for industrial trucks under time pressure, so the reports often contain abbreviations and colloquial wording that make them hard to read, translate, correct, and consistently classify into recurring fault categories.

Translations used to come once a week, but now they can be downloaded twice a day.
annual savings in the 6-digit range

Reported impact

Business outcome

KION reduced the time and cost of translating and classifying maintenance reports, removed dependence on an external translation agency, enabled more frequent report availability, and freed quality teams to focus more on solving customer problems and making recommendations.

  • Cost Reduction
  • Productivity Improvement

Solution

KION Group built MERLIN as a productively used generative AI tool for maintenance and quality workflows around its industrial trucks. Service reports created after maintenance appointments are stored and processed on the KION Analytics Platform (KAP), whose cloud resources are used to store data and train the machine learning model. In the first phase, MERLIN uses a glossary plus contextual signals such as the truck model, series, and spare part references mentioned in each report to expand KION-specific abbreviations, correct spelling issues, and translate the documentation into English, creating a consistent dataset for downstream analysis. On top of that normalized data, a custom-built classification model searches for patterns and assigns reports to recurring "master problems." Results are uploaded as confidence-scored proposals for manual review, while the wider data lake and Dremio query layer let teams consolidate and pull data from multiple internal sources.

Data flow

Service technicians create maintenance reports after each appointment, and KION stores that data on KAP cloud resources. MERLIN uses a glossary plus truck model, series, and spare-part context to correct and translate the text into English, then a classification model detects patterns and maps reports to master problems. The resulting confidence-scored proposals are uploaded for manual review, while Dremio queries the broader data lake for consolidated analysis.

Solution architecture

3 components · 2 layers
  1. Serving
    • MERLIN Generative AI tool that corrects, translates, and proposes classifications for maintenance service reports.
    • Dremio Self-service query engine used to consolidate and pull data from different systems and sources via the data lake.
  2. Uncategorized
    • KION Analytics Platform (KAP) Provides the cloud resources used to store service-report data and train the machine learning model.

Architecture clues

  • A data lake and Dremio are used to consolidate data from different systems and sources.
  • A glossary is used to distinguish KION-specific abbreviations from spelling errors.
  • Classification results are uploaded with a percentage confidence level for manual review.
  • KION describes the implementation as a combination of generative AI and a custom-built classification model.
  • The solution runs on cloud resources in the KION Analytics Platform operated by KION Group IT.
  • Truck model, series, and spare-part references are used as context to improve translation and correction.

Evidence from the source

"In a second step, the AI is trained to search for patterns and automatically classify the service reports into the so-called 'master problems'."
"In the area of product maintenance and quality, the first productively used generative AI tool MERLIN is revolutionizing the maintenance process of our industrial trucks."
"In the first phase, the AI was trained with the help of a glossary to distinguish KION-specific abbreviations from spelling errors in the service reports and to either write them out completely or correct them and then translate the documentation into English."
"Overall, the tool can be used to achieve annual savings in the 6-digit range."
"The AI also orients itself on the truck model, series or spare part mentioned to improve the translation and correction of the documentation."
"The MERLIN classification is currently being uploaded as a proposal with a percentage confidence level for further manual review."
"The combination of generative AI and a custom-built classification model has shown excellent results"
"The main goal is to reduce the effort required to classify problems that are documented during maintenance processes on our material handling equipment."
"To store the data and train the ML model, the team uses the cloud resources of the KION Analytics Platform (KAP), which is operated by KION Group IT."
"Translations used to come once a week, but now they can be downloaded twice a day."
"With the data lake as the basis and self-service query engines such as Dremio, data from different systems and sources can now be easily consolidated and pulled."
Transparency Evidence & methodology 80/100 reference value · 94% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value80/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth20/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact15/25
    1 metric describe measurable impact.
  • Deployment scale5/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence94%
  • Source authority25/25
    The deployment is described by the adopting organization or a government source.
  • Claim traceability21/25
    10 of 12 key fields link to a named source.
  • Direct evidence20/20
    11 direct source quotes retained.
  • Date provenance10/10
    The publication date comes from the source page or publisher.
  • Source transparency8/10
    1 source reference document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity10/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. KION Group

    KION Group customer story

    First-party
    https://www.kiongroup.com/en/Newsroom/Story-Categories/Innovation/Article/Artificial-intelligence-is-revolutionizing-the-maintenance-process-of-industrial-trucks.html?storyid=53120
    Source type
    Customer-published story
    Published
    Jun 12, 2024
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes
    • Publication date

Date evidence

  • Published on source page Jun 12, 2024 · KION Group · First-party
    Open date source
Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.