Reducing Friction in an AI-Driven Workflow
Alfa is an AI tool for investment professionals that helps generate reports and track market updates. Early versions required users to configure agents before fully understanding the insights, which made the experience feel complex. I helped redesign the workflow into clear stages - allowing users to ask questions, review reports, and automate only when needed. This reduced cognitive load and made the product easier to adopt without removing its power.
My Role
Product Designer (UX, Product Design, User Research)
Stakeholders
CEO and Executive Leadership, Product Managers, Engineers, and Designers
Status
Alfa 2.0 launched successfully
Boosted.ai is an AI-powered investment intelligence platform that helps institutional investors automate research, monitor portfolios, and generate actionable insights through chat and autonomous agents.
After evolving from a static data platform to an AI-driven automation system (Alfa), the company grew ARR from $0.5M to $7M in 9 months, improved early retention from 63% to 89%, and achieved 83% DAU among paid users. The platform now serves 20,000 users, has processed 4M+ queries, and has won enterprise clients including TD Direct Investing, FactSet, and Questrade.
Background
When I joined Boosted.ai, our product, Boosted Insights, was a data-driven dashboard for investment professionals. It allowed users to analyze stocks, compare portfolios, and explore market data in depth.
​
As AI capabilities matured, we realized our dashboard product couldn’t fully support users’ need for personalized analysis, we introduced Alfa 1.0 - a conversational experience that allowed users to ask questions in chat, generate structured research reports, and turn those reports into automated agents that continuously tracked specific companies, themes, or risk factors over time.
It was ambitious.
It felt modern.
It showcased the power of AI.
DESIGN EVOLUTIONS
Over three years at Boosted.ai, I helped shape the company’s core product as it progressed through three key design evolutions.
Boosted Insights
Data focused dashboards
Alfa 1.0
A "3-pane" conversational AI experience
Alfa 2.0
Consumption-First AI Workflow
The Problem
After launching Alfa 1.0, active user rates began to decline, and a clear pattern emerged. Feedback from both sales and users was consistent:
​
​​
​
Ease of use became a recurring loss driver.
​
Within the product, we observed:
​
-
Low completion of agent setup
-
Early drop-off during first sessions
-
Heavy reliance on Customer Success during onboarding
​
Users struggled to:
​
-
Understand where they were in the workflow
-
Know what would happen after sending a prompt
-
Navigate between chat, reports, and agents
​
This led to confusion, lower completion rates, and reduced product adoption.
​
​
“It feels complicated.”​​
What Alfa 1.0 Does
In Alfa 1.0, chat, report generation, and agent configuration all lived on the same screen.
​
Users could:
-
Chat.
-
Report building.
-
Agent configuration.
-
Scheduling.
​
​Users experienced significant cognitive overload - asking questions, interpreting results, structuring outputs, and considering automation simultaneously.​​
DISCOVERY
A Misalignment in Mental Models
I partnered closely with the Product and Design teams to investigate where friction was occurring in the experience. I conducted interviews with 12 investment professionals and analyzed 36 user sessions through FullStory to identify behavioral patterns and drop-off points.Through journey mapping and qualitative research, a clear mismatch emerged.
​
I led live user testing sessions to evaluate usability and identify friction in the Alfa 1.0 experience.
Investment professionals naturally follow a progression:​​​
Users were thinking in a sequence (Chat → Report → Agent), but the interface presented everything at once. This mismatch made the workflow harder to understand and use.
However, Alfa 1.0 required users to configure early on in the process. It merged two distinct mental modes: understanding and building - into a single workflow. ​​​​​​​This overlap increased cognitive load and made the experience feel more complex than it actually was.
INSIGHTS
Separating Mental Modes
Through our research, we identified two distinct cognitive states within the investment professional’s workflow:
= Informed Investment Decisions
Understanding requires focus and clarity
Users need space to interpret insights and evaluate what they are seeing​
Building requires decisions and commitment
users must structure outputs, configure automation, and determine next steps.​​
​​In Alfa 1.0, these two states merged into a single canvas. Users were expected to interpret insights while simultaneously configuring agents and automation. By asking them to understand and commit at the same time, we increased cognitive load and made the experience feel more complex.​
SOLUTION
Designing Alfa 2.0
To address the problem, I restructured the workflow around a simple principle:
​
​
​​
One Task At a Time.
We restructured the experience into clear, sequential stages:
​
-
Chat- Users begin in a dedicated chat page to ask questions and explore ideas.
-
Report - Generated reports open in a separate, focused report page designed for reading and analysis - distinct from the chat interface.
-
Schedule - Users can enable automation by converting the report into an agent that delivers scheduled updates.
-
Follow-up - Users can bring Alfa back to ask additional questions, refine the report, or extract new insights from the agent.
Each step had a distinct purpose, allowing users to move from exploration to automation in a more intentional and focused way. By separating these stages, we reduced early commitment pressure and made the workflow more intuitive and easier to navigate.
Alfa 2.0 Interactions
1. Chat
​
Users begin by asking Alfa a question. From there, they can receive a quick answer or generate a structured report for deeper analysis.
2. Report
The report opens in a dedicated view designed for focused analysis.
​
From this page, users can schedule the report, turning it into an automatically updating agent. This allows the system to rerun the report on a set schedule and continuously deliver updated insights.
3. Chat on the Side
​
Alfa 2.0 allows users to ask follow-up questions or refine the report without disrupting the reading experience.
The work and design for Alfa 2.0 may look simple, but it was shaped through many iterations. I worked through multiple versions to align the flow with a clear mental model and ensure the interface felt intuitive rather than overwhelming.
Product Evolution
BEFORE
Alfa 1.0 - Chat and Report on the Same Page
​​​
Configuring while reading reports
Multiple decisions happening at once
Automation introduced before insight was validated
AFTER
Alfa 2.0 - Chat First, Then Report​
​
Dedicated report page for focused reading
Automation enabled after insight is validated
Scheduled delivery based on user intent
Impact
By separating building & undertanding that separate the 2 views of chat and report, we observed:
My Contribution
As a Product Designer on this project, I helped drive the shift from a combined build interface to a clearer, staged workflow. By analyzing onboarding friction and user behavior, I identified the conflict between understanding and building, and advocated for separating these mental modes.
​
I worked closely with the Product team and collaborated with the two other designers to define the new sequential flow and shape the final Alfa 2.0 experience - from chat to report to automation. This was a structural redesign grounded in how financial professionals actually think and work, aligning the product with their natural progression from question to insight to action.
Reflection
