In-Person: Proactive Insights Copilot – Designing Agentic AI Systems on AWS

Most analytics today are reactive; they wait for someone to ask the right question or check the right dashboard. With Generative AI and Agentic Systems, we can flip this model: building assistants that are proactive, monitoring data signals, detecting anomalies, and pushing actionable insights to decision-makers without waiting for a prompt.

In this workshop, set in the context of an e-commerce company, you’ll learn how to design and implement a Proactive Insights Copilot using AWS services. The copilot combines knowledge grounding with specialized agents and event-driven triggers to deliver context-rich, timely insights that move teams beyond static dashboards.

The workshop is guided by AWS’s latest prescriptive best practices.

• Building Serverless Architectures for Agentic AI on AWS – scalable, event-driven architectures using Amazon Bedrock, EventBridge, and Lambda.
• Agentic AI Frameworks, Protocols, and Tools on AWS – guidance for choosing frameworks, integrating tools, and orchestrating multi-agent systems using protocols such as the Model Context Protocol (MCP).

By the end of the day, you’ll have hands-on experience building a copilot that anticipates needs, surfaces insights automatically, and operates with governance and trust, all demonstrated through an e-commerce lens but applicable across industries.

Who Should Attend

• Solutions Architects exploring Generative AI and agentic design patterns
• Developers & Engineers wanting hands-on experience building with AWS services
• Data & Analytics Leaders looking to extend BI capabilities with AI
• Business & Operations Stakeholders curious about how AI surfaces insights proactively

What You’ll Need

• Basic familiarity with AWS services (S3, Lambda, EventBridge, IAM)
• Comfort navigating the AWS Management Console
• Awareness of business KPIs (e.g., conversion, cart abandonment, Average Order Value)
• No advanced coding or ML expertise required

What Will Be Provided

• Access to a ready-to-use AWS environment for hands-on labs
• Sample e-commerce data for use cases and exercises
• Guided instructions and starter resources (Lambda functions, EventBridge rules, datasets)
• A reference guide and cleanup steps to continue learning post-workshop

Agenda (Full Day, 8:30 am – 4:30 pm)

8:30 – 9:00 | Check-In (30 min)

9:00 – 9:30 | Welcome & Why Proactive Insights (30 min)

• Framing the challenge: why dashboards and traditional BI miss critical signals.
• Introduction to Generative AI, AI Agents, and Agentic Systems.
• Overview of the Proactive Insights Copilot use case.

9:30 – 10:00 | Exploring E-Commerce Data with Generative BI (30 min)

• Use Amazon QuickSight with Amazon Q to explore KPIs in natural language.
• Example KPIs: cart abandonment, Average Order Value (AOV), refund rate, and inventory stock levels.
• Demonstrate how Generative BI reduces dependency on queries and dashboards
  

  10:00 – 10:10 | Morning Break (10 min)

10:10 – 11:40 | Building a Smarter Assistant with Knowledge Grounding (1 hr 30 min)

• Create an Amazon Bedrock Knowledge Base (KB) with KPI definitions, policies, and playbooks.
• Demonstrate retrieval-augmented generation (RAG) for contextual, grounded answers.
• Example: refund spike explained using policy + playbook context.
  

  11:40 – 12:25 | Lunch Break (45 min)

12:25 – 2:50 | From Assistant to Agentic System (2 hr 25 min)

• Move beyond Q&A to agents that plan and act.
• Configure a Bedrock Agent with specialized tools: Sales KPI Checker Agent, Inventory Checker Agent, Refund Anomaly Detector Agent.
• Demonstrate how specialized agents collaborate as part of an agentic system, orchestrated by an Advisor Agent.
• Hands-on build and testing with time for troubleshooting.
• Mini-governance touchpoints: least-privilege IAM roles, audit trails, tool contracts.

2:50 – 2:55 | Afternoon Stretch Break (5 min)

2:55 – 3:35 | Making Insights Proactive (40 min)

• Use Amazon EventBridge to trigger insights automatically.
• Example triggers: weekly KPI briefing, cart abandonment > 50%, inventory below threshold.
• Deliver proactive insights via Slack, Teams, or email.
• Discuss event-driven patterns for scaling proactive insights.

3:35 – 4:05 | Governance, Guardrails & Lifecycle (30 min)

• Why trust and explainability matter for proactive systems.
• Governance (AgentOps): version control with AppConfig, pipelines with CodePipeline, monitoring with CloudWatch, and GuardDuty.

4:05 – 4:30 | Wrap-Up & Q&A (25 min)

• Recap: from BI dashboards → knowledge-grounded assistant → agentic system → proactive insights.
• Participant reflections and discussion of use cases.

Also check out other Workshops in Euless, Contests in Euless.