2024

Qubera: An AI Research Tool for High-Stakes Financial Decisions

Financial analysts were losing hours every night reading documents that a well-designed AI could summarize in seconds. I designed the end-to-end product experience for Qubera — an AI-powered research tool built to cut through the noise and get analysts to the insight faster.

Role: Product Designer, Conversational Designer
Duration: 6 weeks
Scope: User research, information architecture, wireframing, prototyping, design system
Tools: Figma, FigJam, Jira

A laptop screen displaying a business analytics dashboard for Nvidia Corp. with financial data, key performance indicators, and a chat interface with questions about the company.

The Users

Junior to mid-senior financial analysts — the people who do the actual research behind investment decisions. They spend hours, sometimes entire nights, reading through 10-Ks, 10-Qs, and press releases. They're working under real pressure: a company releases earnings, the market moves, and an analyst has a narrow window to read, interpret, and make a call.

What surprised me during research: the bottleneck wasn't intelligence or expertise — it was time and access. Analysts knew what questions to ask. Getting to the answers was what was consuming them.

The Problem

Equity research is built on publicly available information — SEC filings, earnings reports, press releases. But accessing, parsing, and synthesizing that information manually takes 15+ hours per analyst per cycle. By the time an analyst has read everything they need, the window for timely decision-making has often already closed.

The goal: design an AI tool that cuts extraction time dramatically and gets analysts to the questions that actually require their judgment.

The Time Constraint

We were building fast. Financial institutions were already in conversation with the team — during one research interview, a director told us she was waiting for the product to launch because she wanted her analysts using it immediately. That kind of validation accelerated everything.

The design had to be production-ready quickly, which meant making deliberate decisions about what to include in the MVP and what to defer.

What I Designed

Default View — Key Financial Metrics

When an analyst selects a company, the most critical financial numbers are already there — revenue, gross profit, net income, EPS, EBITDA, free cash flow, debt — automatically parsed from 10-Ks and 10-Qs. No manual extraction. No hunting through filings.

This single feature reduced data extraction time by 85%.

Screenshot of a dashboard app called Qubera with information about Nvidia Corp. It displays key performance indicators such as revenue, gross profit, operating income, net income, EPS, EBITDA, free cash flow, and debt, with data for fiscal year 2023 and quarter 1 of 2024. The interface includes a welcome message, a focus input box, and four questions about Nvidia Corp.

Focused Chat

Analysts can query the AI directly, focused on specific documents — SEC filings, press releases, earnings calls. They can type a question or select from pre-generated prompts. The AI responds with sourced, specific answers.

This boosted decision-making speed by 30% — analysts could ask targeted questions and get targeted answers instead of reading entire documents to find one data point.

Screenshot of a data dashboard titled 'Welcome to Qubera' with NVIDIA Corp - NVDA financial metrics and a chat box titled 'I am ready to tell you about Nvidia Corp' with options for security filings and press release.

Intelligent Questions — My Proposal

During research interviews, a pattern emerged: analysts ask largely the same set of questions about every company they cover. Revenue growth, competitive positioning, key risks, guidance. The questions were consistent — only the answers changed.

I proposed the Intelligent Questions feature: when an analyst selects a company, a curated set of pre-answered questions is already waiting for them. No typing, no repetition, no cognitive overhead of deciding what to ask first. More importantly, it shifts where analysts spend their mental energy — routine questions are handled, so they can focus on the complex, high-judgment questions that actually require their expertise.

The MVP delivered a standard set of IQ questions for every company. The roadmap included a future version where analysts could build and save their own personalized question sets — we designed for that extensibility even though we didn't ship it in V1.

Screenshot of a Qubera platform interface showing a discussion about Nvidia Corporation's competitive advantages, with sources and answers, and side notes about revenue streams and customer base.

Press Release Summaries and Sentiment Analysis

New press releases are summarized automatically. Analysts see a sentiment score on a color-coded scale — positive, neutral, negative — without having to read the full release first. If something warrants deeper attention, it's one click away.

Screenshot of a market analysis webpage titled 'Welcome to Qubera,' showing NVIDIA's stock and market trends, with a discussion on disruptive market trends affecting NVIDIA, including climate change concerns, rapid technological changes, geopolitical tensions, cryptocurrency market volatility, and export control restrictions.

Validation

Testing was done with actual clients — financial analysts and institutional users who were evaluating the product for adoption.

Formal documented feedback wasn't captured before the project concluded, but the real-world signal was clear: a director at a financial institution was actively waiting for launch to put her team on it.

Impact

85% reduction in data extraction time

30% faster decision-making for analysts using focused chat

Designed and shipped MVP in 6 weeks

Product positioned for launch with active institutional interest

Reflections

This project taught me how much context matters when designing for expert users. Financial analysts aren't confused about their work — they're constrained by the tools available to them. The design challenge wasn't simplification. It was elimination — removing every step between an analyst and the insight they were already capable of drawing.

The IQ feature is the decision I'm most proud of. It came directly from listening carefully during research, noticing a behavioural pattern, and proposing a solution that addressed it systematically rather than feature by feature.

What I'd do differently: structure client testing more formally from the start. We had real users evaluating the product but no structured feedback capture. That data would have been valuable both for iteration and for telling the story of this project.

The best AI tools don't replace expert judgment — they clear the path to it.