Product Design · Fintech · 2023

Richie AI

Portfolio management, made personal.

Making investing feel as intuitive as checking your messages.

R
Richie AI
Verde Finance
Ask Richie anything…
Product Design · Fintech · 2023

Designing a trusted
AI companion.

"Now part of Aetherum.ai" — A personalised AI investment companion that makes financial decision-making feel intuitive, proactive, and human — built for Verde Finance, now part of Aetherum.ai

RoleDesign Intern
Duration3 Months
PlatformiOS · Android
StatusDeployed
01 — Context

Investing is intimidating.
Richie AI makes it personal.

Most Fintech apps overwhelm users with data. Richie AI needed to do the opposite — surface the right insight at the right moment, in language that feels personal rather than algorithmic.

67%
of users felt overwhelmed by the investment dashboard on first use
43%
dropped off before completing the onboarding flow
2.1×
higher retention when users received a personalised insight in session one
02 — The Insight

The research finding that
changed everything.

My quantitative data showed the what, but it didn't explain the why. After eight user interviews, a clear pattern emerged: Users didn't distrust the AI—they distrusted themselves.

"Most people don't avoid investing because they lack money — they avoid it because they don't trust themselves to make the right call."
— Recurring theme across 6 of 8 user interviews

Every design decision after that moment was filtered through one question: Does this build trust, or erode it?

03 — Process

From user insight to execution.

Four phases. No skipping steps.

01
Discover
8 user interviews, analytics review, session recordings, competitive audit
02
Define
Persona synthesis, journey mapping, HMW framing, trust framework
03
Ideate
Crazy-8s, paper wireframes, design principles for AI transparency
04
Build
High-fidelity screens, design system, component library in Figma
03b — User Flow

Mapping the system before
designing the screens.

Before any screen was designed, the full app flow was mapped — every entry point, every decision, every dead end. The key discovery: Richie Chat needed to be reachable from anywhere in the app, not siloed as a tab.

Richie AI user flow dark

Key flow decision — Richie Chat is accessible from a persistent bottom bar on every screen. Users don't navigate to it — it's always there. This was a deliberate choice to reduce the friction between having a question and asking it.

04 — Research Insights

Three findings.
One clear direction.

Insights drawn from analytics data, session recordings, and user interviews across varying levels of investment experience.

01 Top Finding
Users trusted AI recommendations more when they could see the reasoning — not just the outcome. Opacity killed trust faster than a bad recommendation.
02 Biggest Surprise
Users didn't want more features. They wanted fewer, better-timed moments. A single well-placed insight beat a dashboard full of charts every time.
03 Key Pivot
Onboarding wasn't a form — it was a relationship. Users who felt heard during setup were significantly more likely to trust the AI's first recommendation.
05 — User Personas

Three users.
Three relationships with money.

Every design decision was filtered through these three archetypes — real patterns drawn from our interview sessions and usage data.

AM
Alex M.
Anxious Beginner · Age 26
Background
First job, some savings. Wants to invest but paralysed by fear of making the wrong call. Opens the app, sees charts, closes the app.
Needs
Reassurance. Simple language. A clear first step that doesn't feel risky.
"Just tell me what to do — I don't need to understand all of it yet."
SP
Sarah P.
Busy Professional · Age 34
Background
Earns well, has a portfolio, checks it once a month. Knows she should be more hands-on but doesn't have the time to manage it actively.
Needs
Proactive alerts that matter. No noise. Tell her when she needs to act — not just when something changed.
"I don't want to check daily. I want it to tell me when I need to pay attention."
JK
James K.
Active Investor · Age 41
Background
Experienced. Uses multiple platforms. Switched to Richie AI for the AI layer but frustrated that the interface treats him like a beginner.
Needs
Depth on demand. Full data when he wants it. Explanations that match his knowledge level — not dumbed down.
"I already know what a P/E ratio is. Stop explaining the basics."
06 — The Big Idea

Stop showing users everything.
Let them ask instead.

The core insight from research was simple but uncomfortable: traditional stock dashboards show users everything about a stock — price, volume, P/E ratio, 52-week range — all at once. For most users, that's not information. That's paralysis.

"I just wanted to know if I should be worried about Tesla. Instead I got a page of numbers I didn't understand."
— Interview participant, 5 months into investing

The answer wasn't to simplify the data — it was to change the interaction model entirely. Instead of presenting everything and hoping the user finds what matters, Richie Chat lets users ask exactly what's on their mind and get a direct, conversational answer.

Old Model

Dashboard dumps everything

Open a stock. See 20 data points. Feel overwhelmed. Close the app. Repeat. The information was technically accurate but emotionally useless — users couldn't locate what they actually needed to make a decision.

More data, less confidence.

Richie Chat

Conversation surfaces what matters

User types "Is Apple a good buy right now?" Richie responds with a 3-sentence answer tailored to their risk profile and portfolio — with the option to go deeper. The full data is still there. It just doesn't lead anymore.

One question. One answer. One decision.

07 — Design System

Building the visual language
for Richie Chat.

Every component was designed around one principle: the interface should feel like a knowledgeable friend, not a financial terminal.

Brand
Dark
Success
Loss
Caution
Surface
Richie Chat Interface
The core surface — a conversational input that lets users ask any question about any stock. Richie responds in plain language, personalised to their risk profile. The full data is always available one tap deeper, but it never leads. The question leads.
Adaptive Response Depth
Richie's default answer is always 3 sentences — beginner-friendly, jargon-free. A "Show me more" tap reveals charts, ratios, and full analyst data. This single toggle served Alex, Sarah, and James without ever asking them to declare their experience level.
Conversational Onboarding
Goal-setting and risk tolerance captured through the same chat interface — not a form. Users enter the app already familiar with how Richie talks, because onboarding and the core product feel identical. No mode switch, no learning curve.
Proactive Follow-ups
Richie surfaces unprompted check-ins when something meaningful changes in a stock the user has asked about before. Not a price alert — a contextual nudge: "You asked about Apple last week. Here's what changed." Relevance over frequency.
07b — How It Was Designed

The decisions behind
the screens.

I didn't start with the chat interface. I started with the question: at what moment does a user actually need information about a stock? Not when the app opens — when they have a specific thought. "Should I sell?" "Is this dip temporary?" "What happened today?" That's a question-shaped moment, not a dashboard-shaped moment.

Why conversational?
In every interview, users described their anxiety in the form of a question — never a statement. "I don't know if I should be buying right now." "Is my portfolio too risky?" That language told me the interface needed to match how they were already thinking. A chat input is the most natural container for a question.
Why not just search?
Search returns results. Richie Chat returns an answer — shaped by the user's portfolio, risk tolerance, and history. That personalisation is what makes it feel like a financial advisor rather than a search engine. The underlying data is the same; the framing is completely different.
How I handled depth
The hardest design challenge was serving Alex (anxious beginner) and James (active investor) in the same interface. The solution: Richie's default answer is always beginner-friendly — 3 sentences, plain language, one recommendation. A "Show me more" tap reveals the full data layer. Depth on demand, not depth by default.
The constraint I worked within
This was a 3-month internship with an existing product and engineering team. I couldn't redesign the entire app — I had to identify the highest-leverage intervention. Richie Chat was a feature addition, not a rebuild. That constraint actually sharpened the thinking: one new surface, maximum impact.
08 — Iteration

What broke.
What we fixed.

Round 1
Users didn't know what to ask Richie
The first version launched with a blank chat input. Users froze — a blank prompt felt like an exam question. Added suggested prompts ("Is Tesla a good buy?", "Why is my portfolio down?") that disappear once the user starts typing. Engagement with the chat jumped significantly.
Fixed
Round 2
Richie's answers felt too long and text-heavy
Early responses averaged 6–8 sentences. Users skimmed and missed the key point. Redesigned the response format: one bold headline answer, two supporting sentences, then a "Show me more" tap for depth. Reading time dropped, comprehension improved.
Redesigned
Round 3
Users couldn't find the chat from the home screen
Richie Chat was buried in the navigation in early builds. Users went straight to their portfolio and never discovered it. Moved the chat input to a persistent bar at the bottom of every screen — always one tap away, regardless of where the user is in the app.
Validated
09 — Screens

From wireframe to final design.

Every screen went through at least 3 iterations. Getting the flow right before the pixels.

Richie AI screens dark Richie AI screens light

Replace these wireframe placeholders with final Figma exports — modified for NDA as needed. The layout and labels are already structured to match your screen titles.

10 — Outcomes

Designed for trust.
Measured by retention.

Within the internship period, redesigned flows were shipped to production. Early metrics from A/B testing showed meaningful movement across all three target areas.

41%
AI recommendation action rate — up from 12%
2.1×
Retention lift for users with a personalised session-1 insight
3
Rounds of usability testing across all three personas
11 — Next Time

What I'd do differently
with more time.

01
Test the chat input earlier — much earlier
I validated the concept through interviews before designing, but I wish I'd put a rough prototype of the chat interface in front of users in week two rather than week six. Some of the wording and tone decisions I made late would have been better informed by early conversation testing.
Earlier
02
Map the edge cases for bad or uncertain AI responses
What does Richie say when it genuinely doesn't know? When market conditions are too volatile to give a clear answer? I designed the happy path well. The error states and uncertain states deserved the same rigour — in fintech, those moments are high stakes.
Incomplete
03
Build a conversation history that creates value over time
Richie Chat in its current form is stateless — each conversation starts fresh. With more time, I'd design a history layer where Richie remembers past questions and proactively follows up: "Last month you asked about Tesla's dip — here's what happened since." That's where the real retention lives.
Future