Natural Language Analytics Assistant Using Amazon Bedrock Agents

skai · Media & Entertainment · 2025 · source Agentic: yes · promotional bias: med · source-backed · extracted with prompt 0.5.6 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Advertising analytics

Skai needed customers to access complex omnichannel advertising data without waiting days or weeks for manually generated reports and case studies.

Constraints:

Advertising data comes from multiple sources.
Manual report generation took days or weeks.
The assistant handles sensitive customer data requiring enterprise-grade security and privacy.
Outputs needed to include recommendations, not just raw data retrieval.

Architecture

Skai built Celeste with Amazon Bedrock Agents so customers can ask natural-language analytics questions, trigger data collection across sources, synthesize insights, and receive recommendations.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  celeste-assistant["Celeste Assistant"]
  amazon-bedrock-agents["Amazon Bedrock Agents"]
  advertising-data-sources["Advertising Data Sources"]
  insight-synthesis["Insight Synthesis"]
  enterprise-security-controls["Enterprise Security Controls"]
  celeste-assistant --> amazon-bedrock-agents
  amazon-bedrock-agents --> advertising-data-sources
  amazon-bedrock-agents --> insight-synthesis
  advertising-data-sources --> amazon-bedrock-agents
  insight-synthesis --> celeste-assistant
  enterprise-security-controls --> celeste-assistant
  enterprise-security-controls --> advertising-data-sources

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`Celeste Assistant`	Receives customer analytics questions and returns synthesized recommendations.	`Amazon Bedrock Agents`
`Amazon Bedrock Agents`	Orchestrates analytics data collection and reasoning steps.	`Advertising Data Sources`, `Insight Synthesis`
`Advertising Data Sources`	Provide campaign, performance, and customer analytics inputs.	`Amazon Bedrock Agents`
`Insight Synthesis`	Turns retrieved data into reports, case studies, and recommendations.	`Celeste Assistant`
`Enterprise Security Controls`	Protect sensitive customer advertising data.	`Celeste Assistant`, `Advertising Data Sources`

Data flow

A customer asks Celeste a natural-language question, Bedrock Agents orchestrate collection from multiple advertising data sources, the assistant synthesizes insights, and it returns recommendations and report-style outputs.

“enabling natural language queries that automatically collect data from multiple sources, synthesize insights, and provide actionable recommendations.”summary, stated

“Skai, an omnichannel advertising platform, developed Celeste, an AI agent powered by Amazon Bedrock Agents”summary, stated

Why this architecture

Bedrock Agents fit the need for multi-source collection and recommendation synthesis rather than a simple single-query chatbot.

“automatically collect data from multiple sources, synthesize insights, and provide actionable recommendations.”summary, stated

“maintaining enterprise-grade security and privacy for sensitive customer data.”summary, stated

Models used: Amazon Bedrock Agents

Scale

Report generation time fell by 50%.
“reduced report generation time by 50%”summary, stated
Case study creation time fell by 75%.
“case study creation by 75%”summary, stated
Some weeks-long analytics processes became minute-scale.
“transformed weeks-long processes into minutes”summary, stated

Alternatives considered (roads not taken)

Manual report generation

Rejected because: It took days or weeks and delayed customer analytics work.

“time-consuming manual report generation (taking days or weeks)”summary, stated

Single-source analytics chatbot

Rejected because: The use case required collecting data across multiple advertising data sources.

“automatically collect data from multiple sources”summary, stated

Patterns

`tool-use-orchestration` (agent-architecture)

Bedrock Agents orchestrate data collection and synthesis across advertising sources.

“an AI agent powered by Amazon Bedrock Agents”summary, stated

`safety-policy-enforcement` (governance)

Enterprise security and privacy constraints are part of the assistant design for sensitive customer data.

“maintaining enterprise-grade security and privacy for sensitive customer data.”summary, stated

Tradeoffs

Use an agentic analytics assistant

Chose Bedrock Agents-powered Celeste over Static reports and manual analyst workflows.

Rationale: Agentic orchestration reduced reporting and case-study turnaround substantially.

“The implementation reduced report generation time by 50%, case study creation by 75%”summary, stated

Return recommendations, not only metrics

Chose Insight synthesis and actionable recommendations over Raw data retrieval alone.

Rationale: The customer workflow needed decisions and recommendations from complex ad data.

“synthesize insights, and provide actionable recommendations.”summary, stated

Gotchas (what bit them)

Analytics assistants need to synthesize across sources, not just answer from one table.

Root cause: Omnichannel advertising analytics spans multiple data sources and report formats.

Resolution: Use Bedrock Agents to collect, synthesize, and recommend across sources.

“automatically collect data from multiple sources, synthesize insights, and provide actionable recommendations.”summary, stated

Customer analytics data raises privacy and security requirements.

Root cause: The assistant operates on sensitive customer advertising data.

Resolution: Keep enterprise-grade security and privacy controls in the architecture.

“enterprise-grade security and privacy for sensitive customer data.”summary, stated

Evaluation

Skai measured workflow time reduction for report generation, case studies, and weeks-long processes.

Metrics: Report generation time reduction · Case study creation time reduction · Process turnaround time

“reduced report generation time by 50%, case study creation by 75%, and transformed weeks-long processes into minutes”summary, stated

Outcomes

Report generation time dropped by 50%.
Case study creation time dropped by 75%.
Some weeks-long workflows became minute-scale.

The case does not disclose user counts, but it reports large workflow-time reductions for customer-facing advertising analytics.

“transformed weeks-long processes into minutes while maintaining enterprise-grade security and privacy for sensitive customer data.”summary, stated