Unifying Customer Intelligence: Transforming Sales Operations with CAKE and Amazon Bedrock

Introduction

Building cohesive and unified customer intelligence across your organization starts with reducing the friction your sales representatives face when toggling between Salesforce, support tickets, and Amazon Redshift. A sales representative preparing for a customer meeting might spend hours clicking through several different dashboards—product recommendations, engagement metrics, revenue analytics, etc.—before developing a complete picture of the customer’s situation.

Solution Overview

We built the Customer Agent & Knowledge Engine (CAKE), a customer-centric chat agent using Amazon Bedrock AgentCore to solve this challenge. CAKE coordinates specialized retriever tools—querying knowledge graphs in Amazon Neptune, metrics in Amazon DynamoDB, documents in Amazon OpenSearch Service, and external market data using a web search API.

Why Customer Intelligence Systems Need Unification

As sales organizations scale globally, they often face critical challenges, including fragmented data requiring hours to gather comprehensive customer views, loss of business semantics in databases, and manual consolidation processes.

Agent Framework Design

Our multi-agent system leverages the AWS Strands Agents framework to deliver structured reasoning capabilities while maintaining the enterprise controls required for regulatory compliance and predictable performance.

Building the Knowledge Graph Foundation

CAKE’s knowledge graph in Neptune captures customer relationships, product usage patterns, and industry dynamics in a structured format, enabling efficient reasoning.

Engineering for Production: Reliability and Accuracy

When transitioning CAKE from prototype to production, we implemented critical engineering practices for reliability, accuracy, and trust in AI-generated insights.

Results and Impact

CAKE has transformed how AWS sales teams access and act on customer intelligence, reducing the time spent searching for information from hours to seconds.

Conclusion

In this post, we highlighted how Amazon Bedrock AgentCore supports CAKE’s multi-agent architecture, enabling teams to focus on domain expertise and customer value rather than complex infrastructure development.

Unifying Customer Intelligence: The CAKE Solution

Building cohesive and unified customer intelligence is critical for any organization seeking to enhance sales effectiveness and customer engagement. One of the most pressing challenges for sales teams—especially as they scale globally—is the friction created by toggling between various platforms like Salesforce, support tickets, and complex data warehouses such as Amazon Redshift. As sales representatives prepare for meetings, they often spend hours navigating multiple dashboards to piece together a complete customer profile.

At AWS, we experienced this challenge firsthand as we expanded globally. To address this issue, we developed the Customer Agent & Knowledge Engine (CAKE)—an intelligent chat agent designed to bridge siloed customer data across various metrics databases and external industry sources, all without requiring the heavy lifting of custom orchestration infrastructure.

What is CAKE?

CAKE harnesses the power of Amazon Bedrock AgentCore to streamline data retrieval and provide meaningful insights through natural language queries. By coordinating specialized tools such as knowledge graphs in Amazon Neptune, metrics in Amazon DynamoDB, documents stored in Amazon OpenSearch Service, and external market data via a web search API, CAKE delivers actionable customer insights in under 10 seconds.

Key Features of CAKE:

• Dynamic Intent Analysis: Understands customer queries and intelligently routes them to the most appropriate tools.
• Parallel Execution: Executes tool calls concurrently, vastly reducing the time to insights.
• Row-Level Security: Enforces security and governance policies within workflows.
• Production Engineering Practices: Ensures reliability and adherence to business standards.
• Performance Optimization: Provides flexibility in model selection for improved efficiency.

Why is Unification Essential?

As sales organizations grow, they face significant hurdles:

1. Fragmented Data: Sales representatives often toil over disparate dashboards, gathering fragmented insights across various tools—each reinforcing silos instead of providing a cohesive picture.
2. Loss of Business Semantics: Traditional databases may fail to capture the complex interrelationships that explain customer behavior and metric relevance.
3. Manual Consolidation: Processes that work for smaller data sets do not scale, leading to inefficient practices that impede timely decision-making.

A unified customer intelligence system is crucial for overcoming these challenges. CAKE stands as an essential linchpin for enterprises, streamlining data aggregation while ensuring context-rich understanding and actionable insights.

Solving the Problem: The Architecture of CAKE

CAKE reshapes data interaction by transforming disconnected data into an actionable conversational interface. Unlike conventional tools that merely report metrics, CAKE’s semantic foundation allows it to unravel the underlying connections between different data points, thus elucidating not just "what" is happening but "why" it matters and "how" to respond.

The Technical Backbone

CAKE operates on Amazon Bedrock AgentCore, with its custom agent coordinating five specialized tools optimized for distinct data access patterns:

• Neptune Tool: Queries relationship data efficiently.
• DynamoDB Tool: Enables rapid metric lookups.
• OpenSearch Tool: Conducts semantic searches across unstructured data.
• Web Search Tool: Gathers market intelligence.
• Row-Level Security Tool: Guarantees secure data access.

Workflow Overview

Here’s how CAKE effectively processes queries:

1. Intent Analysis: The supervisor agent evaluates the customer’s question and determines the required tools.
2. Parallel Dispatch: Multiple tools are activated simultaneously, taking advantage of Amazon Bedrock AgentCore’s scalable architecture.
3. Result Synthesis: As results are returned, CAKE synthesizes them into a comprehensive, coherent answer.
4. Security Enforcement: Security boundaries are respected throughout the data retrieval process, ensuring compliance and data governance.

This approach minimizes latency and maximizes the relevance of insights provided to the users.

Knowledge Graph: The Heart of CAKE

CAKE’s knowledge graph, built using Amazon Neptune, is pivotal for understanding and reasoning about customer relationships, product usage patterns, and industry dynamics. Instead of compartmentalizing information, the knowledge graph captures semantic relationships, allowing intelligent agents to query and reason more effectively.

Relationship Context

Each entity in the knowledge graph—from customers to products—holds contextual attributes that enhance understanding. For instance, when analyzing customer interaction, attributes like engagement rates and business drivers become critical for interpreting behaviors and formulating strategic recommendations.

Transforming Insights and Results

CAKE has fundamentally altered the way AWS sales teams access and leverage customer intelligence. By condensing the time spent searching for data from hours to mere seconds, sales representatives can now dedicate their energy to strategic customer engagement rather than data collection.

The Impact

Feedback has shown that CAKE significantly reduces the friction traditionally faced by sales teams, enabling them to generate insights from multiple data sources concurrently. The result is a more comprehensive understanding of customer relationships and opportunities.

Conclusion

In summary, CAKE embodies the future of customer intelligence, illustrating how integrated multi-agent systems can streamline the sales process. Built on Amazon Bedrock AgentCore, CAKE enables teams to focus on delivering value rather than on the complexities that often accompany infrastructure.

For organizations looking to innovate their sales strategies and elevate their customer engagement, the unification of data through tailored solutions like CAKE may just be the key to unlocking unprecedented success.

With the rapid evolution of AI and data analytics, tools like CAKE highlight the importance of leveraging technology to create actionable intelligence. For further exploration of Amazon Bedrock AgentCore and building efficient multi-agent systems, we invite you to refer to the Amazon Bedrock User Guide and explore available resources.

Acknowledgments

We extend our sincere gratitude to our leadership team and dedicated members who made this initiative possible, highlighting the exceptional collaboration that has driven our innovations in customer intelligence.

By harnessing advanced AI technologies, organizations can unify their customer insights, refine their approaches, and ultimately drive better business outcomes.