The client
Chatwit uses QueryDash, our agentic investigation platform, for sales analytics.
The challenge
Chatwit wanted its users to investigate business and sales performance without waiting for a technical team to prepare and run database queries. The old way meant finding the right data, writing SQL, running several queries and interpreting the results before reaching a conclusion.
- Data investigation and analysis were manual.
- Users depended on technical staff to write SQL queries.
- Complex business questions needed many queries to answer.
- Findings from different data points were hard to connect.
- There was little room for follow-up questions during an analysis.
- Moving from a business question to an answer you could act on took too long.
How the agent works
Understand the goal. The agent restates what you are trying to find out.
Plan. It breaks the question into steps.
Discover the data. It finds the tables and columns that matter, and how they relate.
Generate SQL. It writes the SQL for the next step.
Validate: read-only. Every statement is checked before it runs. Writes and schema changes are rejected.
Execute. The query runs against production data, read-only.
Analyse. It reads the results and decides what they mean.
Replan or retry. If a query fails or the answer is incomplete, it adjusts the plan and tries again, until it can explain the answer.
What we did
- 01Discovery
We worked with Chatwit to understand what conversational analytics had to do, and where an AI agent could run multi-step investigations on its own.
- Understood the conversational analytics requirements
- Identified the relevant business data and database structures
- Mapped how the available data relates
- Picked out complex, multi-step investigation scenarios
- Defined secure, read-only access to the production database
- Designed how the agent plans, executes, reasons and recovers
- Found where manual data-analysis effort could be cut
From this we designed QueryDash, our agentic investigation platform, on Amazon Bedrock with Amazon Nova Pro.
- 02Onboarding
We onboarded Chatwit onto QueryDash, so its users can investigate business performance through natural-language conversations. The agent interprets the goal, works out what it needs, discovers the database structures, writes and runs validated SQL, analyses the results and decides the next step.
- Database connectivity
- Amazon Bedrock and Amazon Nova Pro configuration
- Secure management of database credentials
- Read-only SQL validation
- Conversation context across an investigation
- The multi-step agentic investigation workflow
- Real-time streaming of responses
- Deployment of the application on AWS
- Testing against representative investigation scenarios
- 03Operations & support
We run and support the solution and its AWS environment day to day, keeping the application available, database access secure and the agent reliable.
- Monitoring of the application and AWS infrastructure
- Monitoring of how the AI agent runs
- Database connectivity support
- Query execution and validation support
- Troubleshooting investigation workflows
- Fixing production issues
- Application logging
- Secure credential and access management
- AWS infrastructure maintenance
Production access is controlled in the application: every SQL statement the agent writes is checked before it runs, and write or schema-changing operations are rejected.
- 04Optimisation & advisory
We keep working with Chatwit to make the agent more effective, reliable and scalable.
- Tuning the conversational investigation workflows
- Better agent planning and multi-step reasoning
- Better SQL generation and validation
- Stronger error recovery and replanning
- Better follow-up conversations
- Reviewing AWS performance and utilisation
- Monitoring application and agent performance
- Reviewing security and access controls
- Advising on new conversational analytics scenarios
- Finding further tasks AI can automate
Because the architecture is reusable, Chatwit can add new investigation scenarios without building a separate analytics workflow for each one.
The architecture
- Amazon Bedrock
- Nova Pro
- ECS Fargate
- Aurora PostgreSQL
- DynamoDB
- Secrets Manager
The result
Chatwit’s users now investigate business and sales performance by asking questions in plain language. They don’t need to know SQL or depend on fixed dashboards. The agent works through a goal step by step: it finds the relevant data, writes and validates queries, analyses the results, weighs the evidence, recovers from failed queries and carries on from what it finds. Amazon Bedrock, Amazon Nova Pro, AWS compute, managed databases, conversation memory, real-time streaming and read-only data access together make it a production-ready agentic analytics service.
Users ask in natural language instead of writing SQL or waiting for a technical team.
The agent runs multi-step investigations, recovers from failed queries and shows evidence for its findings.
Follow-up questions keep the context of the investigation.
Production data stays protected with read-only, validated access.
The same platform extends to new questions without building a new analytics workflow each time.
Key benefits
- Business questions answered in conversation
- Multi-step analysis the agent runs on its own
- Plain-language questions, no SQL
- Finds the right data by itself
- Follow-up questions keep their context
- Findings backed by evidence
- Recovers and replans when a query fails
- Results streamed in real time
- Read-only, controlled production access
- Less manual SQL analysis
Want something like this for your team?
Tell us the questions your team asks or the work that slows you down. An engineer will suggest a small first build on Amazon Bedrock.
