What challenges did AWS SMGS face before AI implementation?

Complex business insights were trapped in static dashboards and manual reports. Leaders had to reconcile fragmented data across multiple systems and wait for curated reports. The organization needed secure, on-demand BI with role-based access and better agility.

What was the impact of the AI solution?

Deployed to more than 4,000 active users. Reduced business review preparation time from hours to minutes. Improved debugging efficiency from tens of minutes or hours down to single-digit minutes. Increased leader confidence through validated, grounded responses and access controls.

Rank#789of 3,449

Source

ProductionEvidence: Medium50/100

NarrateAI conversational business intelligence for AWS SMGS using Amazon Bedrock AgentCore

Use case typeConversational analyticsUpdated May 27, 2026

AWS leaders manage complex data across multiple hierarchies while making time-sensitive decisions that impact global operations. NarrateAI delivers on-demand, context-rich business intelligence to leaders across AWS SMGS through a conversational interface powered by Amazon Bedrock AgentCore.

Organization: AWS SMGS
Industry: Tech & Comms
Location: United States
Published: May 2026

Reported outcomes

4,000 users

active usersAdoption & scale

−90%troubleshooting time

Strategic outcomes

Better decisions & insightEnabled on-demand conversational business intelligenceRisk & complianceImproved security and access control with row-level isolation and guardrailsEmployee experienceIncreased leader confidence in data-driven decisions

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

Normalized claim

Active users: 4,000 users increase

AWS Machine Learning BlogMay 27, 2026Blog postExplicit claimMedium evidence strength

More than 4,000 active users

Normalized claim

Troubleshooting time: 90% decrease

AWS Machine Learning BlogMay 27, 2026Blog postInferred claimMedium evidence strength

Troubleshooting time dropped from tens of minutes or hours down to single-digit minutes.

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

Customer: AWS SMGS
Provider: AWS
Maturity: Production
Linked source: AWS Machine Learning Blog

Deployed to more than 4,000 active users

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 2 of 2

1Conversational analytics
2Executive analytics

Complex business insights were trapped in static dashboards and manual reports.
Leaders had to reconcile fragmented data across multiple systems and wait for curated reports.
The organization needed secure, on-demand BI with role-based access and better agility.

Built NarrateAI with a two-layer architecture that separates batch persona-based narrative generation from real-time conversational interaction.
Used Amazon Bedrock AgentCore to orchestrate specialized agents for intent routing, persona retrieval, response grounding, and validation.
Stored persona-specific narrative files in Amazon S3 and used Amazon Redshift and AWS Lambda to process data, with Amazon Bedrock Guardrails and observability for safety and production monitoring.

Technologies

Amazon Bedrock AgentCore Amazon Bedrock Amazon Quick Suite Amazon S3 Amazon Redshift AWS Lambda Amazon CloudWatch Amazon DynamoDB Amazon Bedrock Guardrails

Deployed to more than 4,000 active users.
Reduced business review preparation time from hours to minutes.
Improved debugging efficiency from tens of minutes or hours down to single-digit minutes.
Increased leader confidence through validated, grounded responses and access controls.

Architecture

NarrateAI uses a two-layer design. A batch knowledge-engine path applies configuration-driven SQL templates, transforms results into structured JSON, renders persona-based narratives with Jinja templates, and stores role-scoped text files in Amazon S3 with row-level security. A real-time conversational path uses Amazon Bedrock AgentCore Runtime to orchestrate specialized agents that detect intent and complexity, retrieve the relevant persona narrative sections, ground and validate answers, and respond with Claude Sonnet 4 via Amazon Bedrock. The stack also includes Amazon Redshift, AWS Lambda, Amazon CloudWatch, Amazon DynamoDB (for prior custom memory before migration), Amazon Quick Suite, and Bedrock Guardrails.

Sources & evidence1

Evidence: Medium50/100Evidence strength