HCLTech Data Analyst Assistant – Agentic AI for Data Analysis

HCLTech Data Analyst Assistant is an agentic AI solution that lets business users, analysts, and data teams explore, analyze, visualize, and govern enterprise data by asking questions in plain English. Instead of waiting for hand-written SQL or searching disconnected catalogs, users get trusted answers, interactive charts, and presentation-ready reports in a single conversation.

HCLTech's professional services team implements, integrates, and tailors the solution in your AWS environment and enables your teams to use it.

• Natural-language analytics - Ask in plain English; the assistant turns questions into SQL or Python over Amazon Athena, Amazon Redshift, Amazon S3, and the AWS Glue Data Catalog, as well as sources such as Snowflake, PostgreSQL, MySQL, and Amazon Aurora. Every answer shows the query it ran.
• Data discovery and profiling - Finds the right tables, explains schemas, and summarizes data quality including nulls, duplicates, and freshness.
• KPI and metric Q&A - Answers use consistent business definitions rather than ad hoc interpretation.
• Visualization and dashboards - Generates charts and dashboards, with export or hand-off to Amazon QuickSight.
• Insight narration - Trend summaries, variance and "why" analysis, anomaly flags, and simple forecasting.
• Deep dives - Cohorts, segmentation, and root-cause drill-down.
• Report generation - Excel, PowerPoint, or PDF output, plus scheduled briefs.
• Data preparation - Helps with cleaning and joins, and combines uploaded CSV, Excel, and PDF files with warehouse data.
• Data quality and catalog publishing - Generates quality rules or accepts them in plain English from a CSV, shows pass/fail scorecards, and publishes approved checks and Amazon S3 file metadata to your data catalog.

• Identity-aware access - The agent inherits the user's permissions (row- and column-level security, AWS Lake Formation, AWS IAM Identity Center). It never uses a super-user service account.
• Explainability - Every answer shows its SQL, source tables, filters, and assumptions.
• Guardrails - Read-only by default, PII masking, prompt-injection defenses, and query cost and runtime limits.
• Auditability - Full logs of who asked what, which data was touched, and what was returned. Human approval is required for any write or action.
• Accuracy evaluation - A golden question set per customer, accuracy scoring, and regression tests on every change.
• Certified answers and feedback loop - Analysts can approve or correct answers, and the corrections improve the semantic layer.
• Data handling - Data stays in your account and region, with no training on customer data and a choice of models.

Users ask a question in a web app. An AI agent determines which data is needed, finds it through your data catalog, runs the query on the user's behalf with the user's own access, and returns the answer with its query, charts, and reports. It remembers the conversation so follow-up questions build on earlier answers, and it adapts its depth to the user's role. The solution runs in your AWS account, and you choose the language model, including models on Amazon Bedrock.

The solution leverages Amazon CloudFront, AWS WAF, Amazon Cognito, Amazon S3, Application Load Balancer, Amazon EC2 Auto Scaling, Amazon RDS for PostgreSQL, Amazon Bedrock, AWS Secrets Manager, VPC endpoints (AWS PrivateLink), NAT gateway, AWS IAM, AWS KMS, AWS Certificate Manager, Amazon CloudWatch, and AWS CloudTrail.

Source systems include Snowflake, PostgreSQL, MySQL, Amazon Aurora, Amazon Redshift, Amazon Athena, and Amazon S3, with other systems integrated as needed. Data catalog options include OpenMetadata and Zeenea today, with Collibra, Alation, Unity Catalog, Amazon SageMaker Catalog, Amazon DataZone, or AWS Glue Data Catalog available for integration.

HCLTech deploys the solution on AWS end to end - connecting and configuring your data sources and data catalog, setting up roles and the data visible to each role, configuring the language model and guardrails, tailoring prompts, business definitions, and tools to your data domain, and training your analysts, stewards, and engineers so they are productive quickly.

Data and analytics teams that want to get from question to insight without hand-written SQL, data stewards who need to profile and govern data without custom scripting, and data engineers who want one conversational workbench across cloud data sources.