What challenges did Hapag-Lloyd face before AI implementation?

Customer feedback analysis was largely manual and reactive. Reviewing hundreds of ratings and comments could take hours or days. Teams needed faster, scalable insights for product planning and feature prioritization.

What was the impact of the AI solution?

Processes over 15,000 feedback items per month. Achieves about 95% sentiment classification accuracy on a labeled test dataset. Reduces structured summary creation from hours or days to seconds. Enables decisions within days rather than weeks. Feedback-driven actions increased positive comments and reduced negative feedback.

Source

Evidence: Low35/100

Hapag-Lloyd uses Amazon Bedrock to transform customer feedback into actionable insights

Use case typeCustomer experience analyticsUpdated Jun 13, 2026

Hapag-Lloyd, a global liner shipping company, built a generative AI feedback analysis solution to replace a manual and reactive review process for customer ratings and comments. The pipeline ingests feedback on a scheduled basis from Amazon S3, uses Amazon Bedrock to classify sentiment and generate summaries and themes, and stores results in Amazon OpenSearch Service for search and exploration. Product managers and other stakeholders use Amazon OpenSearch Dashboards and an internal chatbot to drill into trends, ask natural-language questions, and receive concise insights for product planning and prioritization. The solution also uses Amazon Bedrock Guardrails, AWS Lambda, Amazon ECS, AWS CloudFormation, Amazon CloudWatch, AWS CloudTrail, Amazon SES, LangChain, LangGraph, and cross-region inference to support safety, orchestration, notifications, and resilience.

Organization: Hapag-Lloyd
Industry: Logistics
Location: Germany
Published: May 2026

Reported outcomes

95%

timeTime & speed

Strategic outcomes

Better decisions & insightEnabled faster product planning decisionsSpeed & agilityReduced summary creation to secondsScale & capacityAutomated large-scale feedback processingCustomer experience & trustImproved feedback-driven customer sentiment

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Time: 95%

AWS Machine Learning BlogMay 5, 2026Blog postInferred claimLow evidence strength

Achieves about 95% sentiment classification accuracy on a labeled test dataset.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Hapag-Lloyd
Provider: AWS
Maturity: Unknown
Linked source: AWS Machine Learning Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 4

1Customer Feedback Analysis
2Sentiment Analysis
3Knowledge Discovery

Customer feedback analysis was largely manual and reactive.
Reviewing hundreds of ratings and comments could take hours or days.
Teams needed faster, scalable insights for product planning and feature prioritization.

Built a generative AI-powered feedback processing pipeline on AWS.
Scheduled AWS Lambda jobs ingest new feedback from Amazon S3.
Used Amazon Bedrock to classify feedback sentiment and generate summaries and themes.
Indexed processed records in Amazon OpenSearch Service for search, vector retrieval, and dashboards.
Provided an internal chatbot over the OpenSearch knowledge base for natural-language queries.
Applied Amazon Bedrock Guardrails and LangChain/LangGraph orchestration for safety and multi-step workflows.
Used AWS CloudFormation, Amazon CloudWatch, AWS CloudTrail, Amazon SES, Amazon ECS, and cross-region inference to support deployment, observability, notifications, and resilience.

Technologies

Amazon Bedrock Amazon Bedrock Guardrails AWS Lambda Amazon S3 Amazon OpenSearch Service Amazon OpenSearch Dashboards AWS CloudFormation Amazon CloudWatch AWS CloudTrail Amazon SES Amazon ECS LangChain LangGraph Cross-Region Inference

Processes over 15,000 feedback items per month.
Achieves about 95% sentiment classification accuracy on a labeled test dataset.
Reduces structured summary creation from hours or days to seconds.
Enables decisions within days rather than weeks.
Feedback-driven actions increased positive comments and reduced negative feedback.

Architecture

The solution uses AWS Lambda to fetch new feedback from Amazon S3, Amazon Bedrock for sentiment classification and summarization, and Amazon OpenSearch Service as the search and vector store. Amazon OpenSearch Dashboards provides interactive analysis, while an internal chatbot queries the OpenSearch knowledge base. Amazon Bedrock Guardrails are applied for content safety and prompt-injection defense, with LangChain and LangGraph orchestrating multi-step workflows. AWS CloudFormation, Amazon CloudWatch, AWS CloudTrail, Amazon SES, Amazon ECS, and cross-region inference support deployment, monitoring, notifications, runtime hosting, and resilience.

Sources & evidence1

Evidence: Low35/100Evidence strength