Call center ai · Production

Infosys Topaz uses Amazon Bedrock to cut technical help desk call handling time by 60%

The problem

A large energy supplier's technical help desk handled roughly 5,000 calls per week from meter technicians in the field, with average handling times exceeding 5 minutes for the top 10 issue categories—representing over 60% of call volume—and 60–70% of issues being repetitive. Scaling the support team was costly and not sustainable.

Workflow diagram · grounded in source
1
Meter technician calls help desk
Trigger
Meter technicians call support agents from the technical help desk when they cannot fix issues on-site by themselves.
▾ source quote
“Sometimes they call support agents from the technical help desk to get guidance and support to fix issues that they can't fix by themselves”
2
Transcripts stored in S3
Integration
Calls are recorded and transcripts are stored in JSON format in an Amazon S3 bucket.
▾ source quote
“Calls to the technical help desk are recorded for quality and analysis purposes, and the transcripts are stored in JSON format in an AWS Simple Storage Service (Amazon S3) bucket”
3
LLM classifies transcript relevance
Ai action
Claude Sonnet on Amazon Bedrock uses zero-shot chain-of-thought prompting to summarize each conversation and classify it as relevant or irrelevant.
▾ source quote
“the Lambda function uses Anthropic's Claude Sonnet model on Amazon Bedrock. It uses zero-shot chain-of-thought prompting, to first summarize the conversation and then to determine the relevance. If the conversation is disconnected or disjointed (because of signal disturbances or other …”
4
Embed and store relevant conversations
Integration
Shortlisted conversations are chunked, embeddings are generated, and stored in an Amazon OpenSearch Serverless vector store.
▾ source quote
“The shortlisted conversations are chunked, and embeddings are generated and stored in an Amazon OpenSearch Serverless vector store”
5
RAG retrieval for agent query
Ai action
When a support agent submits a query, the help desk fetches the top similar items from the knowledge base.
▾ source quote
“After the user submits a query, the technical help desk fetches the top similar items from the knowledge base”
6
Virtual assistant responds to agent
Output
A virtual assistant built on top of the knowledge base assists the support agent with resolutions.
▾ source quote
“A virtual assistant is then built on top of the knowledge base that will assist the support agent”
7
Agent feedback recorded
Feedback loop
Agents provide like or dislike feedback on each response, which is persisted in a DynamoDB table.
▾ source quote
“Users can provide feedback by choosing either the like or dislike button for each response received”
Reported outcome

The AI assistant now handles 70% of previously human-managed calls, average handling time for the top 10 categories dropped from over 5 minutes to under 2 minutes (a 60% improvement), issues requiring human intervention fell from 30–40% to 20% within the first 6 months, and customer satisfaction scores increased by 30%.

Reported metrics
Weekly call volume5,000 per week
Share of calls in top 10 categoriesover 60%
Repetitive issues share60–70%
AI-handled call share70%
Show all 10 reported metrics
weekly call volume5,000 per week
share of calls in top 10 categoriesover 60%
repetitive issues share60–70%
AI-handled call share70%
average handling time improvement60% improvement
average handling time (before)over 5 minutes
average handling time (after)under 2 minutes
issues requiring human intervention (before)30–40%
issues requiring human intervention (after, within 6 months)20%
customer satisfaction score increase30%
Reported stack
Amazon BedrockAnthropic's Claude SonnetAWS Step FunctionsAmazon DynamoDBAmazon OpenSearch ServerlessAWS LambdaAmazon Titan Text EmbeddingsStreamlitPandasAWS Secrets Manager
◆ Does this fit your context?

Compare to your context

Tell us your scale, team, and constraints. We'll show what changes at your size, what fails at your scale, and whether this case is a fit, needs adaptation, or won't scale to you. Free demo, no signup.

Compare to your context →
~30 seconds · free
Source
https://aws.amazon.com/blogs/machine-learning/how-infosys-topaz-leverages-amazon-bedrock-to-transform-technical-help-desk-operations?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The AI assistant now handles 70% of previously human-managed calls, average handling time for the top 10 categories dropped from over 5 minutes to under 2 minutes (a 60% improvement), issues requiring human interventi…

What tools did this team use?

Amazon Bedrock, Anthropic's Claude Sonnet, AWS Step Functions, Amazon DynamoDB, Amazon OpenSearch Serverless, AWS Lambda, Amazon Titan Text Embeddings, Streamlit, Pandas, AWS Secrets Manager.

What results were reported?

Weekly call volume: 5,000 per week; Share of calls in top 10 categories: over 60%; Repetitive issues share: 60–70%; AI-handled call share: 70% (source-reported, not independently verified).

How is this call center ai AI workflow structured?

Meter technician calls help desk → Transcripts stored in S3 → LLM classifies transcript relevance → Embed and store relevant conversations → RAG retrieval for agent query → Virtual assistant responds to agent → Agent feedback recorded.

Related call center ai cases

WHAT TO DO WITH THIS

Now compare it to your context

This case is one data point. Whether its pattern fits you depends on your volumes, your stack, and your exception load — that comparison is the step no case study can do for you.

Browse the Call center ai cases →