The Challenge
Trueestate needed a faster and more accessible way for business teams to understand data stored across its operational databases. Answering everyday questions about users, properties, activities, products, transactions, and business performance traditionally required knowledge of database structures and SQL, creating dependency on technical teams.
The challenge was to enable non-technical users to ask business questions in natural language and receive accurate, data-driven answers without compromising the security of production databases.
Grey Bracket Devops Automation designed an AI-powered analytics solution that allows users to interact with their data conversationally while enforcing read-only database access and controlled SQL execution.
Business Outcome
The solution transforms data access from a traditional, analyst-dependent workflow into a conversational GenAI experience. Users can ask questions in natural language without needing to know SQL or the underlying database schema.
The AI agent can reason over database structure, select and execute controlled tools, recover from query errors, and translate database results into business-friendly answers.
Discovery
Grey Bracket Devops Automation worked with Trueestate to identify the types of business questions users regularly needed to answer and the data sources involved.
The solution was designed around a natural-language interaction model where users could ask questions such as:
Rather than relying on predefined dashboards, the AI agent dynamically discovers the database schema, examines relevant columns and values, determines the appropriate query, and retrieves the required information.
The architecture supports both MySQL and PostgreSQL, allowing the same AI analytics experience to work across different database technologies.
Onboarding
Grey Bracket Devops Automation implemented a conversational AI data analyst using Amazon Bedrock and Amazon Nova Pro.
The AI agent uses Bedrock's Converse API and tool-use capabilities to investigate data before generating an answer. It can:
- Discover available tables.
- Inspect table structures and columns.
- Explore relevant data values.
- Generate SQL in the correct database dialect.
- Execute the query through a controlled backend tool.
- Analyze returned results.
- Correct its approach when a query fails or produces unexpected results.
- Present the final answer as a table, headline metric, or visualization.
The application was deployed using Amazon ECS with AWS Fargate, while Amazon RDS for MySQL and Amazon Aurora PostgreSQL provide the underlying data stores.
IAM roles were used for AWS service access, eliminating the need for hard-coded AWS credentials or API keys.
Solution Architecture
Operations & Support
The solution was engineered with security and operational controls around the GenAI workflow.
Every SQL statement generated by the AI agent is validated server-side before execution. Only approved read operations such as SELECT, WITH, SHOW, DESCRIBE, and EXPLAIN are permitted. Write, update, delete, schema-changing, and stacked SQL statements are rejected automatically.
Security Controls
- Read-only database access enforced in application code.
- Database credentials stored in AWS Secrets Manager.
- Databases deployed in private VPC subnets.
- No direct internet access to production databases.
Operational Controls
- Automatic limits for unbounded queries.
- Agent execution limited to a controlled number of iterations.
- Conversation context stored in Amazon DynamoDB with a 24-hour TTL.
- Real-time agent progress streamed through WebSockets.
These controls allow users to see the AI's investigation process while maintaining control over what the agent can execute against production data.
Optimisation & Advisory
Grey Bracket Devops Automation optimized the platform to make GenAI-driven analytics practical for day-to-day business use.
- Amazon Bedrock ConverseStream enables token-level streaming so users can see progress without waiting for the entire agent workflow to finish.
- Amazon Nova Pro is configured with deterministic decoding for tool-use scenarios, helping maintain consistent SQL generation.
- The platform dynamically selects the appropriate Bedrock inference profile based on the deployment region, supporting regional deployment requirements.
- The reusable query-agent was packaged independently from the application server, allowing the GenAI capability to be reused across different applications and database environments.
- The architecture supports both MySQL and PostgreSQL and database-specific SQL dialects.
Technology Stack
Summary
Grey Bracket Devops Automation transformed Trueestate's data-access experience from a traditional, analyst-dependent workflow into a conversational GenAI experience.
Users can now ask business questions in natural language and receive data-driven answers without knowing SQL or the underlying database schema.
The solution combines Amazon Bedrock, Amazon Nova Pro, agentic tool use, Amazon ECS Fargate, Amazon RDS, Amazon Aurora PostgreSQL, Amazon DynamoDB, Amazon CloudFront, Amazon S3, AWS Secrets Manager, Amazon ECR, and Amazon VPC to create a secure and scalable natural-language analytics platform.
The key GenAI capability is the AI agent's ability to reason over database structure, select and execute appropriate tools, recover from query errors, and translate raw database results into business-friendly answers.
About Grey Bracket
Grey Bracket is a cloud and DevOps technology company specializing in designing, modernizing, and operating cloud-native solutions.
Our expertise spans AWS cloud architecture, DevOps, application modernization, AI/ML, Generative AI, containerization, automation, security, and managed cloud services.
We help organizations move from traditional technology environments to secure, scalable, and intelligent cloud solutions that deliver measurable business value.