Understanding the challenge
The firm's advisors were spending their most valuable hours on the least valuable work. Before every client review, an advisor or associate pulled account data from the portfolio system, wrote up performance summaries, researched holdings and market moves, and updated the CRM by hand, often two or three hours of prep for a single meeting. "I became a financial planner to sit across from clients, not to assemble binders the night before," one senior advisor said. The firm had watched other companies roll out AI and wanted the same leverage, but in a regulated business the fear was real: a generic chatbot that might invent a number or say the wrong thing to a client was a compliance incident waiting to happen. So the firm did nothing, and its advisors kept losing evenings to prep while their capacity to take on new households stalled.
Our approach
The firm did not need a chatbot that answers questions. It needed agents that do the work, safely, inside the systems advisors already use. We designed a set of AI agents, each with a defined purpose, clear boundaries, and human approval on anything client-facing:
• A research and analysis agent that pulls each client's holdings and the relevant market context and drafts a plain-language review summary, connected directly to the firm's Salesforce CRM.
• A reporting agent that assembles the client review pack from live account data, so advisors start from a finished draft instead of a blank page.
• A knowledge-base agent that answers advisors' internal questions on the firm's policies and compliance rules in seconds, from the firm's own documented sources.
• Human approval points, escalation rules, and monitoring on every client-facing output, so the team stays in control and nothing reaches a client without an advisor's sign-off.
Technical innovation
Each agent has one clear job and a hard boundary around it, and none of them can send anything to a client on its own. The research and reporting agents work directly against the firm's Salesforce records and portfolio data rather than generic prompts, so every figure traces back to a real source, and each draft is queued for an advisor to review, edit, and approve. Because the agents support multi-step work rather than one-off answers, an advisor can hand off an entire review-prep sequence and get back a finished, sourced draft, with the compliance guardrails built into the workflow instead of bolted on afterward.
Outcome
Preparing for a client review went from two or three hours to under one, a 68% cut, and advisors moved that time back to where it earns the firm money, in front of clients. Over the first two quarters advisors took on more households without adding headcount, review packs became more consistent because they came from one sourced template, and not a single client-facing output went out without an advisor's approval, which kept compliance comfortable. "The agent does the prep I used to dread, and I just check it and add the human judgment," the senior advisor said. "I have my evenings back and more time for clients."
