BGL: Agentic text-to-SQL analytics using Claude Agent SDK on Amazon Bedrock AgentCore
BGL Corporate Solutions is a finance and compliance software company serving more than 12,700 businesses across 15 countries. The company needed faster, more accurate natural-language analytics and reporting without creating a bottleneck for data teams, since traditional text-to-SQL approaches were inconsistent. BGL built a production AI agent using Claude Agent SDK hosted on Amazon Bedrock AgentCore. The agent interprets business questions, identifies the right pre-built analytic tables, generates guardrailed SQL SELECT queries, runs Athena queries, writes and executes Python code against CSV results, and returns insights and visualizations in Slack. AgentCore provides stateful, isolated execution sessions to support security and compliance requirements for financial services.
- Organization
- BGL Corporate Solutions
- Industry
- Finance
- Location
- Australia
- Published
- February 2026
Reported outcomes
Strategic outcomes
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- BGL Corporate Solutions
- Provider
- AWS
- Maturity
- Unknown
- Linked source
- AWS Machine Learning Blog
No explicit deployment-stage evidence found.
Primary read
Use case focus
Showing 3 of 4
- 1Agentic analytics
- 2Text to SQL
- 3Business intelligence
- Business users without technical knowledge relied on data teams for queries, creating a time-consuming bottleneck.
- Traditional text-to-SQL solutions did not provide consistent and accurate results.
- Analytics logic across hundreds of domain tables needed to remain governed and aligned to a single source of truth.
- BGL built a production AI agent using Claude Agent SDK on Amazon Bedrock AgentCore.
- The agent focuses on interpreting natural-language questions, selecting pre-built analytic tables, generating SQL SELECT queries, executing Athena queries, and writing Python code to process CSV outputs and create visualizations.
- The implementation uses domain-specific agent skills and project context files so the right product-line knowledge is loaded on demand.
- AgentCore hosts the agent in isolated stateful sessions, supporting session persistence for up to eight hours and identity-based access controls.
- More than 200 employees can validate hypotheses instantly.
- Compliance teams can spot risk trends without learning SQL.
- Customer success managers can retrieve account-specific analytics in real time during client calls.
- Analytics moved from a bottleneck to a competitive advantage.
Architecture
BGL uses Athena and dbt to pre-build analytic tables from its data platform. A user asks a question in Slack; the Claude Agent SDK agent running on Amazon Bedrock AgentCore identifies relevant tables, generates only SQL SELECT queries, queries Athena, stores results in S3, downloads CSVs into the AgentCore file system, runs Python code for further analysis or charting, and returns the output back to Slack. AgentCore provides isolated microVM-based execution sessions with state persistence.
Sources & evidence1
- Customer explicitly identified
- Technical implementation details available
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?
Community
Comments
Loading comments...