Customer support · Production

Hexagon builds HxGN Alix AI assistant for Enterprise Asset Management using AWS generative AI services

The problem

Hexagon's EAM users had to navigate extensive PDF manuals to find information, while base language models lacked product-specific knowledge and frequently produced hallucinations, making them unsuitable without domain grounding.

First attempt

A proof of concept using an off-the-shelf NeMo model without internal knowledge bases revealed hallucination problems and insufficient product knowledge, making RAG integration necessary for accurate responses.

Workflow diagram · grounded in source
1
User submits free-text query
trigger
“user interactions are in free-text format and users might input content including personally identifiable information (PII)”
2
LLM query reformulation
ai_action
“The LLM first identifies the user's intent and generates a more complete query for the knowledge base. Although this requires an additional LLM call, it yields highly relevant results, keeping the final prompt concise.”
3
RAG document retrieval
integration
“Amazon Bedrock Knowledge Bases enhances Alix's ability to retrieve relevant information from stored documents, improving response accuracy”
4
Guardrails safety validation
validation
“we use a combination of specialized prompts along with contextual grounding checks from Amazon Bedrock Guardrails”
5
Answer delivered to user
output
“Provide users with quick, accurate answers, alleviating the need to navigate extensive PDF manuals”
Reported outcome

HxGN Alix was successfully launched, optimizing internal workflows and elevating user satisfaction within secure enterprise environments.

Reported metrics
LLM response latency range1–60 seconds depending on the user's query
Hallucination reductionreducing hallucinations
Internal workflow optimizationoptimized internal workflows
User satisfaction improvementelevated user satisfaction
Reported stack
Amazon EKSMistral NeMoAmazon Bedrock GuardrailsAmazon Bedrock Knowledge BasesAmazon BedrockNVIDIA L40S GPUs
Source
https://aws.amazon.com/blogs/machine-learning/how-hexagon-built-an-ai-assistant-using-aws-generative-ai-services?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

HxGN Alix was successfully launched, optimizing internal workflows and elevating user satisfaction within secure enterprise environments.

What tools did this team use?

Amazon EKS, Mistral NeMo, Amazon Bedrock Guardrails, Amazon Bedrock Knowledge Bases, Amazon Bedrock, NVIDIA L40S GPUs.

What results were reported?

LLM response latency range: 1–60 seconds depending on the user's query; Hallucination reduction: reducing hallucinations; Internal workflow optimization: optimized internal workflows; User satisfaction improvement: elevated user satisfaction (source-reported, not independently verified).

What failed first in this deployment?

A proof of concept using an off-the-shelf NeMo model without internal knowledge bases revealed hallucination problems and insufficient product knowledge, making RAG integration necessary for accurate responses.

How is this customer support AI workflow structured?

User submits free-text query → LLM query reformulation → RAG document retrieval → Guardrails safety validation → Answer delivered to user.

Related customer support cases