Transforming CRM Insights with AI: How TP ICAP Developed ClientIQ Using Amazon Bedrock
This title captures the project’s essence, highlights the innovative technology, and emphasizes the impact on CRM insights.
Transforming CRM Insights at TP ICAP: The Power of AI with Amazon Bedrock
In today’s rapidly evolving business landscape, the ability to extract actionable insights from vast datasets is paramount. This post, co-written with Ross Ashworth from TP ICAP, explores how their Innovation Lab harnessed the power of Amazon Bedrock to revolutionize the way they interact with thousands of vendor meeting records in their CRM. The result? ClientIQ—an AI-powered solution that turns hours of manual analysis into instantaneous insights.
The Challenge: Underutilized CRM Data
TP ICAP faced a significant challenge: tens of thousands of vendor meeting notes accumulated over years in their CRM, rich with qualitative insights but underutilized. Business users were spending excessive time sifting through records, well aware of the valuable information contained within but unable to locate it efficiently. The need was clear: a solution was required to make this information both accessible and actionable.
Introducing ClientIQ: A Custom CRM Assistant
Enter ClientIQ, TP ICAP’s custom AI-driven CRM assistant, which allows users to interact with their Salesforce meeting data via natural language queries. Imagine asking questions like:
- “How can we improve our relationship with customers?”
- “What do our clients think about our solution?”
- “How were our clients impacted by Brexit?”
With capabilities to refine queries, filter responses by time period, and access source documents directly, ClientIQ transforms the user experience. It provides comprehensive responses while maintaining traceability with links back to the original data.
Behind the Scenes: How ClientIQ Works
ClientIQ uses a large language model (LLM) to analyze user queries and direct them through one of two workflows:
- Retrieval-Augmented Generation (RAG) for insights from unstructured meeting notes.
- Text-to-SQL generation for structured data queries.
Notably, ClientIQ respects existing permission boundaries, ensuring users can only access the data they’re authorized for.
A Custom Solution Powered by Amazon Bedrock
The Innovation Lab opted to develop a tailored solution rather than relying on their CRM’s built-in AI assistant. Partnering with AWS, they built an enterprise-grade solution using Amazon Bedrock, enabling them to select optimal models for their specific tasks and deliver a production-ready RAG solution in a matter of weeks.
Key Components of the Solution
-
Amazon Bedrock Foundation Models: Offering a range of models accessible through a unified API, TP ICAP evaluated different foundation models for varying tasks to find the best fit for latency, performance, and cost.
-
Amazon Bedrock Knowledge Bases: This capability enabled the RAG framework, enhancing AI responses with relevant organizational data while maintaining source attribution.
-
Amazon Bedrock Evaluations: Implementing automated evaluations helped ensure that ClientIQ consistently delivered high-quality responses, tracking various metrics for performance assessment.
Innovative Data Ingestion and Query Handling
To ensure ClientIQ remains current, TP ICAP developed a custom connector for seamless data ingestion from Salesforce. By running daily updates and refining data with a custom chunking approach, they optimized retrieval quality, allowing users to retrieve information with precise context.
Strengthening Retrieval and Filtering
ClientIQ uses a hybrid search approach that combines semantic and keyword searches, significantly improving the relevance and context of retrieved documents. This capability allows users to fine-tune their search results further for optimal insights.
Security and Governance
To ensure adherence to Salesforce’s permission framework, TP ICAP implemented a robust security model. By utilizing Okta group claims mapped to specific divisions and regions, ClientIQ provides users filtered access in alignment with their permissions. This ensures data governance remains a top priority.
Evaluation and Continuous Improvement
The Innovation Lab developed a comprehensive evaluation strategy that includes a systematic approach for measuring performance. By integrating Amazon Bedrock Evaluations into their CI/CD pipeline, they continuously validate that any updates maintain high response quality.
Business Outcomes
ClientIQ has transformed TP ICAP’s operations, leading to a staggering 75% reduction in time spent on research tasks. Stakeholders report enhanced insight quality, paving the way for further development into an intelligent virtual assistant capable of tackling more complex tasks across multiple enterprise systems.
Conclusion
Through the ambitious development of ClientIQ, TP ICAP has turned underutilized CRM data into a strategic asset, showcasing the potential of AI solutions to deliver tangible business value rapidly. To explore how you can leverage Amazon Bedrock for similar implementations, dive into the Amazon Bedrock documentation or check out real-world success stories on the AWS Financial Services Blog.
About the Authors
Ross Ashworth
Ross is part of TP ICAP’s AI Innovation Lab, specializing in harnessing Generative AI for a variety of projects. With over a decade of experience in AWS technologies, he is dedicated to developing innovative solutions that deliver business value.
Anastasia Tzeveleka
Anastasia is a Senior Generative AI/ML Specialist Solutions Architect at AWS, guiding organizations in deploying and scaling cutting-edge AI technologies for real-world applications.
This collaboration between technical expertise and innovative solution-building illustrates a path toward the future of enterprise AI—unlocking new value and insights that drive business success.