Democratizing business intelligence: BGL’s journey with Claude Agent SDK and Amazon Bedrock AgentCore

2026-02-03 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

BGL, a leading provider of self-managed superannuation fund (SMSF) administration solutions, partnered with AWS to develop an AI agent for data analysis, addressing common business challenges like reliance on data teams and inconsistent text-to-SQL results. This agent, built using Claude Agent SDK hosted on Amazon Bedrock AgentCore, allows business users to retrieve analytic insights through natural language while adhering to financial services security and compliance. The solution processes complex financial data across over 400 analytics tables, enabling queries like "Which products had the most negative feedback last quarter?" The implementation emphasizes a strong data foundation using Amazon Athena and dbt Labs, separating the AI agent's role to focus on interpreting natural language and generating SQL SELECT queries against pre-structured analytic tables, thereby reducing hallucination and improving consistency, performance, and maintainability.

Key takeaway

For Data Scientists or AI Engineers building natural language data analysis solutions, prioritize establishing a strong, pre-processed data foundation before integrating AI agents. Your agent should focus on interpreting user queries and generating simple SQL against well-structured analytic tables, rather than handling complex data transformations. Leverage modular knowledge architectures and code execution capabilities to manage context efficiently and process large datasets, ensuring consistency and scalability in production environments.

Key insights

A robust data foundation and modular AI agent design are crucial for secure, scalable, and accurate natural language data analysis.

Principles

• Separate AI agent interpretation from complex data logic.
• Organize domain knowledge modularly for efficiency.
• Use code execution for large dataset processing.

Method

BGL's method involves pre-building analytic tables with Athena and dbt, using Claude Agent SDK for natural language interpretation and SQL generation, and executing Python code for result processing and visualization within Amazon Bedrock AgentCore's isolated sessions.

In practice

• Implement `CLAUDE.md` for global project context.
• Define `SKILL.md` files for product-specific domain expertise.
• Host agents on stateful, isolated infrastructure like AgentCore.

Topics

• AI Agents
• Text-to-SQL
• Amazon Bedrock AgentCore
• Claude Agent SDK
• Data Foundation

Code references

• anthropics/claude-agent-sdk-python

Best for: Machine Learning Engineer, AI Engineer, Data Scientist

Related on AIssential

• Build AI agents for business intelligence with Amazon Bedrock AgentCore — AI agents on Amazon Bedrock AgentCore can autonomously process fragmented business data for real-time BI.
• Claude Code for Data Science Projects — Claude Code's core engineering primitives surprisingly provide a strong fit for managing messy data science workflows.
• How OpenAI Built Its Data Agent — The reliability of AI agents at scale stems from strong data foundations and carefully engineered context, not complex agent architectures.
• Claude Opus 4.8 is now available on AWS — Claude Opus 4.8 offers enhanced autonomy and consistency for complex, multi-stage AI workflows on AWS.
• Claude Agent SDK: The Beginner-Friendly Roadmap to Building Production-Ready AI Agents — AI agents represent a paradigm shift from simple chat to autonomous systems capable of planning, tool use, and task execution.
• How We Built an AI Agent for Self-Service Analytics — Effective NL2SQL for business analytics requires structured data and a semantic layer, not just connecting an LLMs to a database.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.