AI Integration for Meeting Preparation

The manual process of preparing for a client review is a classic data aggregation problem. An advisor or associate must log into Addepar for performance and holdings, check the CRM (like Salesforce Financial Services Cloud) for recent notes and activities, open the financial planning software (e.g., eMoney) for goal progress, and scour email or a research portal for relevant market commentary. An AI integration stitches these systems together through their APIs, using a scheduled or trigger-based agent to compile a unified pre-meeting dossier. The agent's core tasks are: 1. Query Addepar's Portfolio and Performance APIs for the period's returns, top/bottom performers, and allocation drift. 2. Pull the client's recent activities, open tasks, and life event flags from the CRM. 3. Retrieve the latest plan values and goal tracking metrics from the planning platform. 4. Synthesize recent firm research or economic updates via a connected RAG system.

The output is not a raw data dump. The AI structures this information into a formatted packet with clear sections: Performance Summary (with plain-language commentary on drivers), Planning Check-In (highlighting progress or deviations), Recent Interactions & Follow-Ups, and Suggested Talking Points. The talking points are generated by cross-referencing data anomalies—like a significant sector overweight or a missed savings target—with the client's stated goals and risk profile. This shifts the advisor's prep time from compiling data to interpreting insights and strategizing. For the firm, it ensures every meeting, regardless of the advisor's tenure or workload, starts from a consistent, high-quality information baseline.

Rollout requires mapping the firm's specific system landscape—custodian feeds into Addepar, CRM custom objects, planning software links—and defining the packet's template and approval workflow. Governance is critical: the AI's data access must respect existing CRM security roles (e.g., only pulling notes the advisor is permitted to see), and all generated content should be flagged as AI-drafted for compliance. A phased implementation often starts with a read-only "assistant" that creates a draft packet for the advisor to review and edit within their existing workflow, building trust before moving to more automated publishing directly to the client portal or meeting software.

The core agent is triggered by a calendar event from the advisor's Microsoft 365 or Google Workspace. It first authenticates with the firm's Identity and Access Management (IAM) platform (e.g., Okta) to obtain scoped API tokens. Using the client name from the meeting subject, it queries the Wealth Management CRM (e.g., Salesforce Financial Services Cloud) to retrieve the client's household ID, recent notes, and open action items. This ID is then used to fetch the client's consolidated portfolio view, performance data, and recent transactions from the portfolio management system (Addepar, Orion, or Black Diamond).

Concurrently, a separate workflow queries the financial planning software (e.g., eMoney) for the client's most recent plan, goal progress, and cash flow updates. A RAG (Retrieval-Augmented Generation) agent, powered by a vector database like Pinecone, searches the firm's internal knowledge base—including recent market commentaries, research reports, and compliance bulletins—for relevant talking points. The orchestration layer (using a platform like n8n or CrewAI) manages these parallel data fetches, handles API rate limits, and merges the results into a structured JSON payload.

The final compilation step involves a generative AI model (e.g., GPT-4, Claude 3) prompted with the structured data. It drafts a concise narrative summary covering performance highlights, planning updates, and recommended discussion topics, ensuring all data is cited to its source system for auditability. The completed packet—containing the narrative, key data points, and source links—is then posted back to the CRM as a pre-meeting note, attached to the calendar event, and/or sent via a secure channel to the advisor's tablet application. Governance is enforced through role-based access control (RBAC) synced from the IAM platform, and all agent actions are logged to a dedicated audit trail in the firm's SIEM or logging platform.

A production AI agent for meeting prep must operate within the strict data governance and compliance frameworks of wealth management. This means architecting the system to pull data on a read-only basis from core systems like Addepar (for performance), your CRM (for recent activities and notes), and financial planning software (for goal updates). The agent should never write transactional data back to these systems without a human-in-the-loop approval step, such as an advisor review of the generated packet before it's saved to the client record or emailed.

A typical implementation follows a phased rollout: 1) Pilot with a single advisor team, where the agent compiles packets for internal review only, focusing on data accuracy and relevance. 2) Expand to a controlled client segment, such as quarterly reviews for a specific book of business, introducing the packet via the client portal with clear disclosure. 3) Firm-wide deployment with role-based access controls (RBAC) in the AI tool itself, ensuring only assigned advisors can generate packets for their clients. Audit logs must track every packet generation—which models were used, which data sources were queried, and who reviewed the output—to satisfy compliance reviews.

The governance model should treat the AI-generated packet as a draft for advisor augmentation, not an automated client communication. Advisors remain responsible for the final content. This controlled integration reduces manual prep time from hours to minutes while keeping the advisor firmly in control of the client relationship and all compliance obligations.