Wealth management

Advisor Copilot

An AI workspace for wealth advisors that runs inside Salesforce and Redtail. As its sole product designer, I led the move from chat-only tools to one workspace, then designed five modules on it.

Role
Product designerResearch, strategy, UI
Company
CogniCorAI for wealth management
Timeline
~2.5 yearsOngoing
Reach
6 enterprise firms~1,000 advisors

A 25-second silent video. Before: three stacked modals over the CRM, each gating the next. Then Advisor Copilot, one AI workspace inside the CRM. Every meeting, from prep to follow-up. Forms pre-filled from the CRM and sent for e-signature. Ask about any client and see the sources. A prioritized day across the whole book. Client emails and reports, drafted with context. Plancorp reported 3 to 10 hours a day back to advisor teams, across about 60 advisors.

Problem

Many tools, low adoption

Advisor Copilot had grown into separate AI tools, each with its own navigation. Firms paid per user and per module, so low adoption put renewals at risk.

ClientMeet
ClientCare
ClientIQ
Each with its own navigation
Low adoption
Priced per user and per module
Renewal risk

Role

Sole designer, three review groups

I ran the competitive and domain research myself. Pushback came from three groups: pilot advisors, engineering in design reviews, and the CEO, CTO and PMs.

Each design
Pilot advisorsUsability
EngineeringFeasibility
CEO, CTO, PMsBusiness fit

Before

Every task went through chat

Each tool ran every action through a linear chat, so advisors couldn't skip a step or switch tasks. Demos weren't landing enterprise clients.

The old chat-gated product: three stacked modals over Salesforce, with the Advisor Copilot chat as the way in.
123
  1. 1Three stacked modals: Each one gated the next
  2. 2Needed 90% zoom: To fit a normal laptop
  3. 3Every action ran as chat: Breaking it meant starting over

Pivot

Made chat optional

Leadership wanted a fully chat-driven product. I argued that live chat across this many connected systems was too slow and unreliable, and proposed a workspace panel with chat one click away.

After: one Advisor Copilot workspace docked in the CRM, with chat beside it.
Before: the chat-gated tool, three stacked modals over the CRM.
BeforeAfter

Drag to compare. Same CRM, same advisor.

Enterprise clients began signing after the pivot. The AI improved at the same time, so I don't credit design alone.

Pattern

One layout for every module

Stacked modals would have repeated in every new module. I set one pattern instead: a section rail on the left and a single working surface.

  1. 1Section railPrep, Agenda and Email in one rail
  2. 2One working surfaceThe rail swaps what fills it
  3. 3Module railSame place in every module
A ClientMeet screen built on the pattern: a section rail with Prep, Agenda and Pre-Meeting Email beside one working surface, and the module rail on the right.123

Principle

Advisors verify the AI's work

Advisors answer to compliance, so the AI does the work and the advisor checks it before anything goes out.

  • Lisa's beneficiary designations haven't been updated in 3 years.

    SourcesSalesforce · Beneficiaries

    Show the source

    Every AI answer links to where it came from.

    ClientMeetClientIQ

  • CustodianSalesforce
    Charles Schwab

    Pre-fill, then review

    The system fills from connected data; the advisor checks it.

    ClientCare

  • Meeting summary: Lisa MartinesDecisions and action items, synced to the CRM

    Check it in the inbox

    Summaries arrive by email, so advisors review them without reopening the app.

    ClientMeet

Constraint

Narrow panel, inside the CRM

Copilot's workspace panel is 790px wide inside Salesforce or Redtail, tested on 13- to 14-inch laptops. Stakeholders and pilot advisors wanted a compact default, so every layout fits it and maximizes when the work needs room.

The Advisor Copilot panel embedded inside the Salesforce CRM, docked at the right.
790px
The meetings view, with chat open beside it.

Embedded

Navigation rebuilt inside the panel

Inside the CRM's iframe, the browser's back button and refresh don't reach the panel, and modals can't escape it. I gave the panel its own back button and breadcrumbs, and kept drafts through reloads.

ClientsSara GansPrepClient workDraft savedtakes heighttakes width123
  1. 1Back and breadcrumbsReplaces browser back
  2. 2Module railSwitch modules in place
  3. 3Draft savedWork survives a Salesforce reload
The rebuilt controls take height and width from the same panel as the client work.

ClientMeet

Meeting prep at a glance

The first module I rebuilt, and the most adopted. Before a call, advisors see the client snapshot, AI talking points and last meeting's decisions in one view. Deep financials stay one click away.

The ClientMeet prep view for Lisa Martines: client snapshot, AI talking points, last meeting recap and sources.1234
  1. 1Client snapshot: Key facts and life events
  2. 2Talking points: One click adds a point to the agenda
  3. 3Sources: Where each insight came from
  4. 4Last meeting: Decisions from last time

Three tabs, one per stage

Upcoming, Past and Scheduler match how advisors move through a day, so tomorrow's prep and last week's notes are one click apart.

The ClientMeet tabs: Upcoming Meetings, Past / Ongoing Meetings and Scheduler.
Course correction

Recording from any meeting

First: a dedicated recording tabShipped: record from Join

In the pilot, advisors couldn't find the tab. I moved recording into the Join button on every meeting, and I later added online and phone recording.

Sentiment Analysis: happiness, frustration and neutral sentiment across the meeting timeline.
Keyword Usage Analytics flagging restricted-word instances with timestamps.

Each recording adds sentiment and timestamped compliance flags.

Upcoming
Join Meeting
LMLisa MartinesTomorrow, 10:30 AMAgenda ready
EHEvelyn HayesFri, 11:00 AMAgenda pending
LHLiam HarperMon, 4:15 PMAgenda pending
Prep: snapshot, talking points, join

ClientCare

Forms arrive mostly filled

ClientCare runs servicing work such as opening an account. The advisor adds what's missing and sends it for e-signature.

Pick a workflow to watch it fill.

Pick a servicing workflow

4 fields pre-filled from connected data · 2 need you
Client
Date of birth
Tax ID
Custodian
Initial contribution
Primary beneficiary
Course correction

Chat builder to form

Scoped: an agentic chat builderShipped: forms to check

Advisors check every field at once instead of rereading a chat.

One builder, then a library

Coding each firm's workflow by hand wouldn't scale, so I designed a builder. Firms share most fields, so a library lets an ops team change a common rule once.

The ClientCare workflow builder: numbered sections, field settings, conditional logic and data mapping.
Sections, fields, conditional logic and CRM mapping, configured without code.

Admin became review

Before

Forms filled field by field, often after the call.

After

Reviewed and sent without leaving Copilot.

ClientIQ

Answers show confidence and sources

ClientIQ answers questions about a client with a generative model, which can be wrong. With the developer, I mapped what the model could produce against what compliance required, then designed each answer to show its confidence and sources.

What should I know before my meeting with Lisa Martines?

ClientIQHigh confidence

Lisa is a 12-year client with a conservative growth strategy. Her last review flagged tax‑loss harvesting before year-end, and her beneficiary designations haven't been updated in 3 years.

Sources
Meeting notes · Aug 15Tasks · Beneficiary reviewSalesforce · Investment strategy

3 similar answers merged into one

  1. 1Confidence on every answerHigh, Medium or Low, and Not scored when there isn't enough to go on.
  2. 2Sources down to the fieldThe Salesforce field, meeting note or statement page each answer used.
  3. 3Repeats mergedModels repeat themselves, so near‑identical output is merged before the advisor sees it.

Platform

Two more modules, same layout

AdvisorIQ covers practice analytics; ClientWrite drafts client emails and reports.

Pick a module to see its screen.

The AdvisorIQ module screen
The whole book: AUM, performance and priorities.

Query your whole book in natural language

Which clients hold ESG?

A prioritized day from all signals

RMD not yet takenCritical
Plan success < 70%High

Satisfaction tracking, then outreach

88On track

System

Architecture became the design system

I designed the architecture first: the panel, the module and section rails, the meeting tabs and the split Join/Record button. A colleague and I then built the design system from it, so all five modules share the same parts.

From the architecture

Module rail

ClientMeet module rail item, activeClientIQ module rail itemClientWrite module rail item

Section rail item

Section rail item, defaultDefaultSection rail item, hoverHoverSection rail item, activeActive

Meeting tabs

Upcoming Meetings tab, activePast Meetings tab

Split button

Split button, primaryPrimary
Split button, secondarySecondary

Controls

Button

Button, primaryPrimary
Button, secondarySecondary
Button, tertiaryTertiary

Foundations

Noto Sans Display

  • heading-10024/32 ExtraBoldLisa Martines
  • heading-50016/22 SemiBoldAgenda
  • default-text14/20 RegularBeneficiaries not updated in 3 years.
  • text-small12/16 RegularTomorrow, 10:30 AM

Color

  • primary-default
  • primary-200
  • text-default
  • text-secondary
  • bg-100
  • bg-200
Components exported from the design system file, with their default labels.

Impact

3–10 team hours back daily

Plancorp measured prep and follow-up time across about 60 advisors and 17 efficiency metrics. Its compliance team approved the figures in January 2026.

3–10hrs/day

Prep and follow-up time recovered per advisor team at Plancorp.

30–45 min
prep reclaimed per meeting
090 min
~60 min
follow-up saved per meeting
090 min
Pilottomulti-year
the lead firm's pilot became a multi-year contract
It felt as though the advisor had known them for years.
Plancorp, on clients meeting a newly assigned advisor

Next

Earlier research, a self-serve tier

  1. My own advisor sessions

    I relied on stakeholders for advisor contact. I'd run sessions across junior, senior, independent and enterprise advisors.

  2. Make the case sooner

    The evidence for the workspace already existed in the chat-only version. I'd put that argument together earlier.

  3. Onboarding in each module

    Onboarding came late. I'd build first-run guidance into each module from the start.

  4. A self-serve tier

    Deep CRM integration keeps enterprise firms but blocks independent advisors. I'd explore a lighter self-serve version for them.

Reflection

The decision I'd defend most is making chat optional. Every module after it was built on that workspace.

Advisor Copilot · Product design case study