PROJECT OVERVIEW
Empowering brokers with AI driven analyst insights
Financial services, AI
A leading brokerage firm needed a way to help it’s brokers quickly access and use analyst research information. They were struggling with too many platforms, unreadable reports, and time-consuming manual tasks.
Our goal was to create an AI Powered Chatbot that would process tons of research PDFs and reports, providing instant, contextual answers with source citations through a conversational AI, and streamlining the dealer's entire workflow into one efficient interface. Ultimately, we aimed to cut down the time it took to find information from minutes to just seconds during critical client calls.
The design phase of this project took a month and a half, and development is currently still in progress. Due to the confidential nature of this project under an NDA, I'm unable to share specific images or detailed visuals of the interface. However, I'd be happy to discuss it on a call.
Timeline
4 weeks for design
Year
2024
Team
Product Designer, Project Managers, Development Team
RESPONSIBILITIES
I was involved in mainly conceptualization and visual design
Research & discovery
I helped conduct user interviews by taking detailed notes and synthesizing research findings to extract actionable insights about dealer workflows and pain points.
Design strategy & conceptualization
Collaborating with another product designer under CEO guidance, I worked on initial conceptualization and strategic wireframing to define the chatbot's core functionality and user experience.
Visual design
I handled the entire end-to-end visual design process of 50+ screens, working closely with the client's established design system to ensure consistency and seamless integration with their existing product ecosystem.
Collaboration & edge cases
I collaborated extensively with both the client team and development team to discover, define, and design comprehensive solutions for edge cases and technical constraints throughout the project timeline.
RESEARCH INSIGHTS
Understanding and observing broker workflows
Our research phase was instrumental in uncovering the root causes of broker inefficiency and directly informed our solution strategy. Through contextual field visits to observe live work environments, in-depth interviews, and detailed workflow analysis, we identified key behavioral patterns and environmental constraints:
Key observations
Keyboard-first operations
Brokers heavily relied on keyboard shortcuts and muscle memory, indicating a preference for fast, efficient input.
Reliance on external tools
Many brokers bypassed internal PDFs, preferring quick insights from external sites like Screener.com.
Juggling with many platforms
They constantly switched between 5-6 different platforms during single client calls, highlighting a fragmented workflow.
Burden of manual tracking
They used Excel sheets to manually track hundreds of client activities and preferences, adding a significant burden to their daily tasks.
THE PROBLEM
Information overload and manual work caused inefficiency
Workflow inefficiency
Brokers faced significant time loss due to information scattered across multiple sources. Constant context switching and platform switching fatigue during critical client interactions also severely reduced their overall productivity.
Information management crisis
Text-heavy PDF research reports were impossible to consume during time-critical client calls. This led to information overload without clear prioritization or summarization, causing decision paralysis from too many unorganized data sources.
Burden of manual process
Managing clients using Excel sheets for 500+ clients was unsustainable. Brokers also faced the burden of manual tracking for client activities, preferences, and interaction history, along with repetitive data entry across multiple systems.
Negative business impact
These pain points directly led to brokers work being inefficient and not meeting their daily transaction goals, which was a key reason the client reached out to us to design and build this platform.
THE SOLUTION
A unified AI chatbot for instant answers
Our initial concept, developed with another designer, focused solely on a chatbot providing AI insights through widgets. However, our goal evolved: we recognized the need to provide brokers with all the information required to complete a transaction within a unified platform. This led to a more comprehensive AI-powered robo-advisory platform that streamlines the entire dealer workflow.
Left Panel — A control hub for actions
This panel acts as a central control hub, offering quick access to research, stock picks, alerts, and other key actions. During discussions with the development team, we decided to move all non-AI-related functions here.
This redefinition allowed the left panel to serve as a complementary action space to the main chat panel, making it both customizable and scalable for any future features.
Central Panel — An interactive AI chat
This is the core interactive space where dealers can query stock insights. The AI draws information from numerous research PDFs and incorporates dealer feedback to continuously improve its responses.
All research provided by analysts is fed into this system, allowing brokers to access insights presented through various dynamic widgets. We designed a comprehensive list of widgets based on the main insights brokers need most while on client calls
Right Panel — Transparency of source
This panel displays the data sources for the AI's responses, offering direct access to the original research files and deeper insights. It's expandable as needed, ensuring it doesn't disrupt the dealer's workflow.
We learned early in our interviews that brokers often send analyst research to clients who request it. To facilitate this, we added citations to every AI message, allowing brokers to easily check and download the relevant PDFs to provide accurate information to their clients.
Overcoming other hurdles during the project
Navigating ambiguous needs
The Challenge
We faced constant uncertainty around content, features, and data accessibility throughout the project. For example, when designing the alert panel, we weren't sure of all the options a broker might need to set an alert.
The Solution
I designed the modal and other compoenents to be flexible, allowing new alert categories to be added that could accommodate additions or removals and designed with modularity in mind to prevent cascading changes. We also established regular feedback loops to manage evolving requirements and ensure that till the end the design did not break the user experience.
Overcoming my lack of domain knowledge
The Challenge
My limited personal knowledge of stock trading and financial markets posed a risk of designing solutions that didn't align with industry practices.
The Solution
I invested personal time in understanding the stock market and trading especially the terminologies that the brokers use.. We also built a strong partnership with the client through daily collaborative and leveraged the client's domain expertise to ensure the deign and experience was built for the end users.
Adapting to aggressive timelines
The Challenge
Our two-person team faced an aggressive four-week deadline to deliver 50+ screens, from concept to final designs, with multiple stakeholder approvals.
The Solution
We segregated our tasks. While the other designer focused on getting approval and working on changes, I prioritized core user states and definable interactions over edge cases in UI. We involved the development team early for feasibility checks and technical validation, and we managed scope through iterative feedback cycles and phased delivery.
METRICS
The metrics that I wished I could have tracked
To validate the success of the new product, I wish I could have tracked key metrics directly aligned with the business objectives.
Workflow optimisation
Did the platform reduce app switching and improve workflow during client queries?
Platform engagement
What were the daily active dealer rates, and were they retained over time?
Information access speed
How quickly could dealers find specific stock insights during live calls?
LEARNINGS
Hard-learned lessons from the project
User-centered design approach
Having the opportunity to visit the client's office and observe dealers in their natural work environment profoundly shaped our design decisions. For example, we learned about their preference for keyboard input over voice commands due to their noisy environment, and the necessity of accessing the chatbot while on client calls.
Design what you can define
Early in the wireframing stage, I realized the challenge of designing for potentially infinite chatbot conversations. Guided by our CEO, I focused on designing core, definable states like the default chatbot view and common widget outputs (charts, tables, lists). This gave us a solid structural foundation before we brainstormed and designed for edge cases with the development team.
Allow buffer time in project planning
This project delivered a crucial lesson in project management. As a two-person design team with a four-week deadline to deliver over 50 screens (from concept to visual design, including wireframes, approvals, and edge cases), we learned the importance of building in extra time—at least a week or two—for feedback cycles and approvals, which often take longer than initially estimated.
Inclusive collaboration from the start
Involving all relevant parties from the beginning proved invaluable. We conducted early feedback and feasibility checks with the development team. This proactive collaboration helped shape the product clearly, minimizing miscommunication and streamlining the development process.