• Wealth advisors struggle with fragmented information, and they need to access multiple systems to correlate and synthesize data. Common systems used by wealth advisors include CRM platforms (e.g., Salesforce, Zoho), Portfolio management tools (e.g., Envestnet, Yodlee), Research platforms (e.g., Yahoo Finance, Morningstar), Financial data providers (e.g., Yodlee, MoneyGuidePro), and various market data sources (Pershing, Broadridge). However, these systems have different user interfaces and data retrieval approaches, which makes it slow, manual, and error-prone for advisors to find the right insight at the right moment.
• This slows down advisor productivity, increases compliance risk, and limits their ability to deliver timely, personalized advice to their clients.

The overall solution is built applying the multi-agent system architecture and Google ADK as the underlying framework & Vertex AI Agent Engine as Cloud Services with key highlights as:

• Agents can autonomously coordinate tasks like retrieving data, summarizing insights, and validating the results as per enterprise-specific guidelines
• Additionally, agents can elevate the conversation with advisors to understand the need better and act accordingly. Reasoning after retrieving data from multiple systems helps to synthesize the information, not just data retrieval.
• Agents can continuously learn from user interactions, remember previous interactions, and use feedback loops to enhance the system incrementally.
• In the future, Agents can also delegate actions to other AI Agents based on the insights discovered, such as updating the CRM system for the next client conversation, reviewing the financial plan, proactively using the interaction to update client communication history, etc.

FA³AI is an agentic knowledge assistant that unifies data and reasoning for wealth advisors. The architecture leverages a multi-agent system (MAS) design using the ADK framework with agents as:

• financial_advisor_agent: Acts as the central intelligence layer, coordinating an ecosystem of specialized agents. It interprets user intent and routes requests to the right agents for knowledge retrieval across multiple systems. Google’s Agent Development Kit (ADK) enables the creation of this centralized orchestrator with built-in reliability, security, and enterprise-grade capabilities. Workflow Agents orchestrator capability helps to manage the execution flow of sub-agents.
• financial_enterprise_docs_agent: Enterprise-specific knowledge retriever agent accessing proprietary, unstructured enterprise financial data sources such as PDFs and CMS-based knowledge hubs to enrich advisor responses with comprehensive document insights.
• crm_leads_insights_agent: CRM Leads and Prospects insights agent integrating CRM data (e.g., Salesforce, Zoho, Redtail, Hubspot) to provide structured, actionable customer relationship and pipeline intelligence in response to advisor queries.
• compliance_checker_agent: Content validation agent ensuring advisor responses are complete, accurate, compliant with legal and regulatory standards, and that any personal or sensitive information is detected and redacted from outputs.

Patterns Applied:

• LLM Orchestrator Pattern for dynamic decision-making by Gemini as LLM
• MCP Integration Pattern for connecting to external, standardized services such as CRM Tools, Third-party Services
• Session Management using Session Service and Persistent Session for enterprises (surviving restarts, failovers)
• Context Compaction to automatically reduce the context that's stored in the Session (future advancement)
• Agent Observability with LoggingPlugin for production systems and ADK Web UI for development and testing
• Agent Evaluation with Google ADK (regression testing using adk eval, Tool trajectory and response metrics)
• Agent Deployment with Vertex AI Agent Engine and long-term memory with Vertex AI Memory Bank
• Agent Communication using A2A protocol as a standardized protocol for agent-to-agent communication (to be implemented for future use as an enterprise expanding solution with external vendors)

If additional time is available, the above solution to be expanded using the construct below. As per the enterprise context, additional agents to be expanded to enrich FA³AI's agentic knowledge assistant, such as:

• portfolio_management_agent: Retrieving and reasoning with enterprise solutions such as Envestnet, Yodlee - each system can be accessed with a specialized associated tool
• financial_research_agent: Researching from the downstream third-party tools for Morningstar, Yahoo Finance, and consolidating the knowledge
• crm_analyzer_agent: Retrieves client's data from CRM systems such as Salesforce, Zoho, Redtail, etc.
• financial_planning_agent: Fetches the client's financial plan (as needed) and use this information to contextualize the information to answer the related questions

graph LR
  %% Actor
  U[Wealth Advisor]

  %% Core
  subgraph Core["FA³AI Agentic System"]
    FA_AG[financial_advisor_agent]
  end

  %% Supporting Agents
  FED[financial_enterprise_docs_agent]
  CRM[crm_leads_insights_agent]
  CC[compliance_checker_agent]
  FA[Future Domain Agents]

  %% Data + Memory
  M[Memory / Knowledge Store]
  DS[Enterprise Data Sources]

  %% Flow
  U -->|query| FA_AG
  FA_AG -->|read/write| M
  FA_AG -->|retrieve docs| FED
  FA_AG -->|retrieve crm data| CRM
  FA_AG -->|delegate| FA

  FED -->|data| DS
  CRM -->|data| DS

  FED -->|results| FA_AG
  CRM -->|results| FA_AG
  FA -->|results| FA_AG

  FA_AG -->|validate| CC
  CC -->|approved| FA_AG

  FA_AG -->|response| U

  %% Styling
  classDef core fill:#cce5ff,stroke:#333;
  classDef mem fill:#fff7cc,stroke:#333;
  classDef ds fill:#f1f1f1,stroke:#333;
  classDef rev fill:#f8d7da,str

• Increased Advisor Efficiency: Higher advisor efficiency by 25-30% (10-12 hours/week efforts saved with seamless access to data from a diverse set of systems)
• Higher Engagement Rate: 10-20% higher engagement rate with clients, with valuable insights enriching personalized engagement
• Quicker Decision Making: Ease of decision making around products, servicing, client portfolio, market insights, etc. based on synthesized information