The Challenge
TNBT had valuable operational data stored in relational databases, but accessing that information often required technical knowledge of database schemas and SQL.
Business users could easily identify the questions they wanted answered, but turning those questions into useful insights required assistance from developers, analysts, or database specialists.
Existing dashboards provided visibility into predefined metrics, but they could not easily support every new or exploratory question. This created a gap between what the business wanted to know and what the existing reporting environment could answer.
TNBT needed a self-service analytics experience that would allow users to explore their data naturally, without requiring them to understand SQL or the underlying database structure.
Grey Bracket DevOps Automation designed a Generative AI-powered data exploration platform that allows users to interact with operational data using natural language while maintaining strict controls over database access.
Discovery
Grey Bracket DevOps Automation worked with TNBT to understand how teams interacted with their existing data and identify the barriers preventing wider self-service analytics adoption.
The solution was designed around a simple principle:
Users should be able to ask questions about their data in the same way they ask questions of a colleague.
Instead of requiring users to know which tables contain the required information, the AI agent dynamically investigates the available data.
Example business questions
- "How many new users joined last month?"
- "Which products are getting the most activity?"
- "Show me the top-performing categories."
- "What changed compared with the previous period?"
- "Which products have the highest demand?"
The agent determines which tables and fields are relevant, investigates the database structure, generates the required SQL, executes the query, and converts the results into an understandable response.
Core requirements
- Natural-language data exploration.
- Self-service access for non-technical users.
- Dynamic schema discovery.
- Autonomous SQL generation.
- Support for multiple database dialects.
- Conversational follow-up questions.
- Automatic recovery from SQL errors.
- Secure read-only database access.
- Real-time visibility into AI processing.
- Support for tables, metrics, and visualizations.
Onboarding
Grey Bracket DevOps Automation implemented a conversational data exploration platform using Amazon Bedrock and Amazon Nova Pro.
The AI agent acts as an intelligent interface between TNBT's users and their underlying databases.
When a user asks a question, the agent determines what information is required and uses controlled database tools to investigate the available data.
Agent workflow
- Understanding the user's question.
- Identifying the information required.
- Discovering relevant database tables.
- Inspecting columns and table structures.
- Exploring actual data values when necessary.
- Generating the appropriate SQL.
- Validating the SQL before execution.
- Executing the approved read-only query.
- Interpreting the returned data.
- Correcting and retrying when a query encounters an error.
- Returning the result in a business-friendly format.
This enables users without SQL expertise to independently explore data that previously required technical assistance.
The platform was implemented using React and TypeScript on the frontend, with a Node.js and Express backend running on Amazon ECS with AWS Fargate.
Amazon S3 and Amazon CloudFront provide delivery of the web application, while Socket.IO enables real-time communication between users and the AI agent.
Solution Architecture
Operations & Support
Because the platform interacts with operational databases, Grey Bracket DevOps Automation implemented application-level controls around the AI-to-database workflow.
The AI model does not receive unrestricted access to the databases.
Instead, all database operations are routed through a controlled execute_sql_query tool.
Every generated SQL statement is validated by the backend before it is executed.
Permitted operations
SELECTWITHSHOWDESCRIBEEXPLAIN
Statements that could modify or destroy data, including INSERT, UPDATE, DELETE, DROP, ALTER, and CREATE, are rejected.
Stacked SQL statements are also blocked, while unbounded queries receive an automatic LIMIT 500 safeguard.
Additional operational controls
- Database credentials stored in AWS Secrets Manager.
- AWS service access controlled through IAM roles.
- Databases deployed in private VPC subnets.
- No direct internet access to production databases.
- Maximum agent iteration limits.
- 24-hour conversation retention using DynamoDB TTL.
- Real-time agent processing through WebSockets.
The result is a self-service analytics environment where users gain greater access to data without giving the AI uncontrolled access to the underlying infrastructure.
Optimisation & Advisory
Grey Bracket DevOps Automation optimized the platform around the goal of making data exploration accessible while controlling the cost and operational impact of GenAI workloads.
Amazon Bedrock ConverseStream provides streaming responses so users can see the agent investigating their question rather than waiting for the complete workflow to finish.
The agent can perform multiple database exploration steps within a controlled execution loop. This allows it to move beyond simple text-to-SQL generation and investigate unfamiliar data structures before producing an answer.
Amazon Nova Pro is configured for deterministic tool-use behavior, helping provide consistent results when the model is generating SQL and selecting database tools.
Conversational follow-up
Short-term conversation memory allows users to continue exploring a topic naturally.
User: "Which category has the highest sales?"
User: "Show me the top five products in that category."
The second question can use the context of the first without requiring the user to repeat the original request.
The reusable query-agent was also separated from the application server, creating a foundation for extending the same data-exploration capability to additional applications and database environments.
Summary
Grey Bracket DevOps Automation helped TNBT move from a traditional reporting model toward self-service, conversational data exploration.
Instead of depending on technical teams for every new data question, users can interact with their operational data using natural language.
The GenAI agent dynamically explores database structures, determines the required information, generates and validates SQL, executes controlled read-only queries, handles errors, and presents the resulting insights as tables, metrics, or visualizations.
The solution combines Amazon Bedrock, Amazon Nova Pro, Amazon ECS Fargate, Amazon RDS, Amazon Aurora PostgreSQL, Amazon DynamoDB, Amazon S3, Amazon CloudFront, Amazon ECR, AWS Secrets Manager, AWS IAM, Amazon VPC, Application Load Balancer, React, TypeScript, Node.js, Express, Socket.IO, and Chart.js.
The key value delivered to TNBT is data democratization: turning operational databases into an accessible conversational knowledge layer for users who do not have SQL or database expertise.
About Grey Bracket DevOps Automation
Grey Bracket DevOps Automation specializes in cloud architecture, DevOps automation, application modernization, Generative AI, AI-powered solutions, cloud security, containerization, automation, and managed cloud services.
We help organizations adopt modern cloud-native technologies to solve practical business challenges and improve operational efficiency.
Our expertise spans AWS cloud architecture, Generative AI, agentic AI solutions, DevOps, application modernization, automation, security, and scalable cloud infrastructure.
We design and implement secure, production-ready solutions that combine cloud, automation, and Generative AI to create measurable business value.