How to build a Investor Helpdesk

Workflow Overview

This flow is designed to help investment managers answer questions about clients and products by combining CRM data and internal documentation, then formatting the result for easy use.

Node-by-Node Breakdown

1. Question (Input Node)

• Purpose:
  This is where the user (investment manager) enters their question.
  Example: “What’s the latest update on Alex Johnson’s account and what estate planning products might be suitable?”

2. CRM Retriever (LLM Node)

• Purpose:
  Uses an Anthropic LLM (Claude Sonnet 4) to search Salesforce CRM and summarize client information.

• How it works:

  • Takes the user’s question as input.

  • Searches the CRM for relevant client details, recent activities, account status, and notes.

  • If information is missing, it politely says so and suggests next steps.

  • Always respects privacy and compliance.

• Special Features:

  • Uses a knowledge base search tool for enhanced retrieval.

  • Connects to Salesforce with a specific connection ID.

3. Product and Services Retriever (LLM Node)

• Purpose:
  Uses another Anthropic LLM to answer questions about the fund’s products, offerings, eligibility, and procedures.

• How it works:

  • Receives the user’s question and the CRM summary as context.

  • Searches internal documentation (like Confluence, product sheets) for relevant product/service information.

  • Summarizes key features, eligibility, and procedures.

  • If information is missing, it suggests where to find more details.

  • Always respects compliance and privacy.

• Special Features:

  • Uses a knowledge base search tool focused on product documentation.

4. Formatter (LLM Node)

• Purpose:
  Uses an OpenAI LLM (GPT-4) to combine and format the outputs from the CRM Retriever and Product/Services Retriever.

• How it works:

  • Takes both the client summary and product/service summary.

  • Formats them into a clear, concise, and professional answer.

  • If both types of information are available, presents them in separate sections.

  • If only one is available, notes what is missing.

  • Ensures the answer is ready for the investment manager to use directly.

5. Answer (Output Node)

• Purpose:
  Displays the final, formatted answer to the user.

How Data Flows

1. User enters a question in the Question node.

2. The CRM Retriever uses the question to pull and summarize client data from Salesforce.

3. The Product and Services Retriever uses the question (and CRM context) to pull relevant product/service info from internal docs.

4. The Formatter combines both summaries into a single, well-structured answer.

5. The Answer node presents the final result to the user.

Key Points

• The flow is fully automated: user asks a question, and the system does the rest.

• It leverages both CRM data and internal documentation for comprehensive answers.

• Formatting ensures the output is professional and actionable.

• Privacy and compliance are respected at every step.