From Idea to Victory: 

The UBS Innovation competition challenged teams to design a tool to enhance wealth advisors' interactions with clients.  

"FinAlly"- The Product which I built leveraged emerging technologies such as Language models, Generative AI Agents, Voice-to-text, and Data aggregation to provide real-time Portfolio analysis, Recommendations, Prompts, and Portfolio Analytics. These features enabled advisors to offer highly personalized and comprehensive financial advice, create a strong impression with clients, and deliver delightful scenarios and “aha” moments during their interactions. 

How I Approached the Challenge

This two-round competition required teams to submit a written proposal in the first round and develop a working prototype in the second if selected. 

Problem Understanding / Empathizing with Users
In the first round, we had limited data, so I started by exploring the UBS website and similar firms website to understand their services and the industry. To gain insights, I reached out to wealth advisors via LinkedIn and cold calls, but most declined or postponed due to their busy schedules. Switching to secondary research, I used Google and Reddit to uncover common frustrations and challenges faced by wealth advisors and their clients. This helped me pinpoint key gaps in the industry. 

After completing secondary research, I identified key industry problems and used prioritization frameworks such as the Eisenhower Matrix and MoSCoW method to rank them by importance and value. I then assessed feasibility using an effort vs. outcome matrix and evaluated technological fit, helping pinpoint the most impactful issues to address in our prototype. 

Wireframe and Digital Prototyping
Next, I began visualizing the best design for web app, identifying key features, and created a wireframe using Balsamiq. I then worked on the software architecture, selecting the required technologies, and assessed the feasibility of implementation. After this, the initial proposal was ready, presenting a robust solution. The process was iterative, refining the design and technology choices as I progressed.

Below is the initial proposal that was submitted:

A few days later, I received the exciting news that my team of two (Robo Rich) was selected for the final round. Now, the challenge was to build a working product based on our proposal and present it to the judges. 

Building the Product 

Next, I began building the product by gathering more technical requirements and sourcing the data needed to fine-tune large language models and financial documents typically used by wealth advisors. During this process, I encountered challenges as conversational data wasn't publicly available due to privacy concerns and GDPR regulations. To overcome this, I generated synthetic data using AI and sourced realistic sample financial statements.

I collaborated with a friend for web development and began building the backend and integrating the tools. We utilized OpenAI’s LLM models, Watson AI, Speech-to-Text, AWS S3 for storage, and Pinecone for the vector database. After a month of effort, we successfully built a working application capable of performing the following tasks: 

1. The application acted as a CRM, storing client financial details, documents, and personal information related to investments and future planning. It was also able to fetch live financial information from banking APIs, providing wealth advisors with a comprehensive and up-to-date view of clients' profiles. 

2. It could summarize and aggregate data from various documents, live accounts, and different platforms, providing accurate information on portfolio allocation. The application generated visualizations to help wealth advisors quickly understand the data, saving valuable time on research. 

3. It had feature, a built-in calling system, allowing wealth advisors to directly connect with clients from the platform. The calls were recorded, and transcripts were stored for future analysis, ensuring seamless communication and easy access to important client interactions. 

4. The Analytics and Advisory AI Engine generated recommendations based on the client's portfolio and conversation history, considering broader context and analyzing market risks before suggesting actions. This feature saved wealth advisors significant time while delivering highly personalized, well-informed advice, leading to delighted clients. 

5. The platform enabled advisors to quickly recall previous conversations and action items, helping them prepare for calls efficiently and gain a deeper understanding of clients' needs. 

6. FinAlly can provide AI-generated call summary with action items, which advisors could instantly email to clients within the platform, saving time and improving communication. 

Check out the final product we created:

Results

We presented our solution to UBS's top leadership, including the CTO, CIO, Chief of Data Science, and Head of Wealth Advisory. They were impressed with the practicality and relevance of our product, praising it for addressing real-world challenges faced by wealth advisors. Our efforts earned us second place in the competition, along with a $5000 reward.