Plum's AI Wealth Advisor screens
Ongoing · Release 1.1 shipped, iterating for Q3 2026

Plum's AI Wealth Advisor

Role
Product Designer, sole designer on this project
Team
1 PM, 1 lead engineer, 1 data analyst, 2 frontend, 3 backend (AI squad)
Timeframe
3 sprints, shipped June 2026, in active iteration

Overview

Plum's chat assistant already helps people save and invest. Plum's AI Wealth Advisor turns that into a full plan.

Pick a goal, like retirement or a house deposit. Get a monthly deposit, a portfolio allocation, and a projection, ready in minutes.

It replaces the cost and intimidation of seeing a human advisor.

The problem

Two problems:

  1. Nobody was finding the advisor. 85% of eligible users never opened it. It lived deep in the app with no clear entry point. Essentially hidden.
  2. Those who found it gave up before getting anything useful. 9 in 10 who started a conversation never saw a plan. The conversations were too long and asked for too much upfront. When a plan finally appeared, it was plain text in the chat, with no way to interact with it.

The result: a 10 to 12% plan save rate and most of the product's potential going untapped.

Goals

Primary goal. Make plan creation a way in for people who feel shut out of financial advice. Many assume it's only for people with more money. A free, trustworthy plan inside an app they already use reaches them directly.

Secondary goal. Move plan save rate from ~12% to a 30% target. Make plans visible, not buried in chat.

Research insights

01

Users abandon before a plan is even generated

People dropped off before a plan even appeared. Too many questions, confusing wording, wrong recommendations, any one was enough to end the flow.

Source: internal plan creation evaluations, qualitative review
02

People will use an AI plan, but only if they can still touch it

Tested automatic vs. editable plans with 10 people. Automatic felt generic and out of their hands. Editable won. Control over deposit and target date mattered most.

Source: Wealth Plans concept test, December 2025, n=10
03

Advice is undersold to the people who need it most

People want reassurance, not bigger returns. Many rule themselves out. They assume advice is only for people with more money. Even past advisor clients wanted a one-off check-in, not a subscription.

Source: Financial Advisor Study, May 2026, n=345 survey plus 8 interviews

Design decisions

Decision 1

Add "Start a Plan" to the homepage's suggested next actions

Why: 85% never opened the advisor. It had no home beyond a buried nav spot. Now it's a suggested next action on the homepage.

Homepage with the new 'Start a Plan' suggested action
Decision 2

Replace the open conversation with a 3-question modal

Why: too many, repeated, unclear questions caused drop-off. Now capped at 3: multiple choice, type-your-own option, no sensitive data upfront.

Decision 3

Generate 3 plan options instead of 1, then open an editable detail screen

Why: a single automatic plan felt generic and out of people's control. Now AI generates a plan in seconds, slow, balanced, or fast-track. Control stays with the user: every option opens into a plan they can edit directly, not one they're locked into.

Decision 4

Add live progress tracking to saved plans

Why: a saved plan on its own doesn't tell you if you're still on track. Now each one shows its status at a glance, on track or behind schedule. It surfaces a concrete next step when it slips. Reassurance becomes ongoing, not a one-time confirmation.

Saved plan detail showing on-track status and suggested action when behind schedule
Decision 5

Run a CRM campaign to people who'd never opened the advisor

Why: no in-product fix reaches people who never see the feature. A targeted campaign drove +140% planning sessions vs. control. 85% were first-time openers.

CRM campaign creative: Plan with Plum

Impact

Sessions nearly doubled post-launch: 3,343 to 11,942.
Plan save rate climbed from a 12% baseline to a 15.2% peak.
Plan Viewed → Saved conversion: 48% to 54.4%.
CRM campaign drove +140% more planning sessions vs. control, with 85% first-time planners.

Honest miss: adoption is still small. Only a few thousand users have a plan. Funded plans outpace saved ones: some skip saving but still complete tasks within 7 days. Plan save rate may be undercounting real impact.

Takeaways

Bring engineers in earlier

There was a real gap between the vision and what we could build. We only found out how big partway through. Next time, engineers see the concept before it's final, not after.

Question the format, not just the flow

I designed around the assumption that people want a goal-first plan. I'd test whether broader coaching is what people actually want, before building around it.

Sequence releases, don't bundle them

We wanted to ship too many things at once. That made it impossible to track the impact of any single decision. Next time I'd push the team to prioritise and stagger releases from the start.