Back

Contextual intelligence and customer support for Durabuilt Windows and Doors

Deveshi Dabbawala

December 9, 2024
Table of contents

Durabuilt Windows and Doors is a leading manufacturer in Western Canada, recognized for operational excellence and consistent innovation. As warranty policies evolved and product lines expanded, the company needed a smarter way to support internal teams handling warranty and product related queries.

Problem: scaling warranty and product support with evolving policies

As Durabuilt updated its warranty policies, support teams faced increasing complexity. Agents manually searched warranty PDFs, compared policy versions, and aligned responses with customer specific product scenarios. This process caused delays, inconsistent answers, and limited scalability as products and policies expanded.

At the same time, manufacturing and lifecycle data remained disconnected from support workflows. Warranty decisions relied on static documents instead of contextual product metadata. Durabuilt needed an AI driven support layer with structured reasoning, contextual classification, and IoT readiness.

Solution: AI customer support chatbot

GoML designed and delivered an MVP AI customer support chatbot to serve as an intelligent warranty and product support assistant. The system provides first level decision support for internal teams and establishes a scalable foundation for connected, IoT-enabled workflows.

The solution leverages an Agentic AI boilerplate. Coordinated AI agents handle query classification, document retrieval, contextual enrichment, and response generation in a structured pipeline.

Conversational warranty and product support assistant

Support teams interact with the AI customer support chatbot through a lightweight interface.

The chatbot:

  • Classifies queries as General, 2023, 2024, or Out of Context
  • Identifies the relevant warranty year logic
  • Retrieves the most relevant indexed policy sections
  • Highlights key differences between policy years when required
  • Generates structured, documentation grounded responses

Vector-based technical documentation retrieval

Warranty PDFs for 2023 and 2024 are stored in Amazon S3 and indexed into OpenSearch.

Amazon Titan Embed Text v2 converts documents and queries into embeddings for semantic search, and Claude 3.5 Sonnet on Amazon Bedrock generates the final response.

This keeps answers grounded in official warranty documentation.

IoT-ready contextual intelligence

AI Matic ingests structured metadata such as product model, manufacturing batch, installation date, and inspection data to align warranty logic with product lifecycle context.

As connected manufacturing systems evolve, IoT signals can link warranty queries to specific production details.  

The chatbot then generates responses based on both policy year and product data, enabling predictive service and data driven warranty analytics.

Backend infrastructure and secure deployment

The MVP includes scalable backend infrastructure provisioned through CloudFormation.

Core components include:

  • Amazon S3 for warranty document storage
  • Amazon OpenSearch for vector indexing and retrieval
  • Amazon SageMaker for execution workflows
  • Amazon Bedrock running Claude 3.5 Sonnet
  • IAM roles for secure access control

Impact

  • 60% Reduction in response time
  • 45% Faster, personalized support
  • 35% Lower support-related operational costs

About

Location 

Edmonton 

Tech stack 

Amazon Web Services, Amazon Bedrock, Claude 3.5 Sonnet, Amazon OpenSearch Service, Amazon S3, Amazon SageMaker 

Before Gen AI and after Gen AI

Area 

Before Gen AI  

After Gen AI 

Support resolution 

Manual PDF search and interpretation 

AI customer support chatbot with semantic retrieval and structured classification 

Policy identification 

Agent dependent and time intensive 

Automated year based query classification 

Documentation access 

Static documents stored in S3 

Vector indexed semantic search with Titan embeddings 

Product context 

Warranty logic disconnected from product lifecycle data 

IoT-ready architecture capable of ingesting manufacturing and product metadata 

“With AI Matic, we transformed warranty support into a structured AI-driven system that connects documentation, policy intelligence, and product context into a scalable support framework.”

Prashanna Rao, Head of Engineering, GoML

Key takeaways for manufacturers

Common challenges

  • Warranty documents grow more complex each year
  • Policy updates increase decision complexity
  • Manual document search slows resolution
  • Product lifecycle data remains siloed from support

Practical guidance

  • Deploy an AI customer support chatbot as the first support layer
  • Use semantic vector search instead of keyword search
  • Implement agentic AI for structured reasoning and workflow control
  • Design systems to ingest IoT and product metadata
  • Log interactions to improve warranty analytics and product insights

Ready to modernize warranty and product support?

Partner with GoML to accelerate the development of production-ready AI customer support chatbots with AI Matic.

Outcomes