Nihal Design Portfolio

RESUME

How I pitched and got buy-in for a CustomGPT at Rocket

My role

Product Strategy, UX Research & Design

Project

Passion project during my internship

Timeline

2 months (July - August 2025)

Tools used

Figma Make, ChatGPT dev tools, Loom

Straight to the pitch

I whipped a prototype of a customGPT using Figma Make and sent a Loom pitch video to the CEO, VP, Senior Product Managers and Designers. The GPT is fine-tuned using Rocket's data and can provide detailed, contextual answers about the home-buying process for users.

They loved the idea!

How this works

RocketGPT is a public-facing, AI-powered assistant designed to serve as a top-of-funnel discovery and lead-generation tool for Rocket.com and RocketMortgage.com. It lives entirely within the ChatGPT ecosystem and answers complex questions about Rocket products before directing users to the websites.

The goal is to extend Rocket’s reach into emerging AI-native user behavior by providing accurate, branded education and seamless pathways back to RocketMortgage.com.

The research was clear — Users weren't happy

I made use of Rocket's existing user research database to find data on the typical Home-buying process. Users were clearly overwhelmed. There were too many questions, very little clarity.

Insights from research

Tailor-made personal experience

The home-buying process is often tough to understand for users. Having a tool that explains concepts in a way that they understand is essential.

User's often quit due to confusion

Find the mortgage calculator

55% of all users quit the application as the they feel it is too long or confusing

User's don't know where to find information

Fill in the details for calculation

Most users get overwhelmed by the website and do not know where to find the information they need.

Client Goals

I identified the top 5 goals clients have when they begin an Online Mortgage Application

User personas — Who are we building this for?

I made personas on different target segments of user groups that would use RocketGPT.

How might we?

There were some crucial tasks we needed RocketGPT to handle for users:

User flow

I sat down with our conversational designer and we charted out the user flow that would allow us to structure RocketGPT's responses.

3 main flows:

Purchase flow - If user intent is Home buying
Refinance flow - If user intent is to refinance existing home
Others - Else if user intent exploring options FAQ, etc

FINAL DESIGN

Calculate your payments

Chat with the GPT to estimate your monthly payment or calculate affordability.

Ask Questions — Get live data

Wondering what the rates are today? Ask the GPT and it will pull accurate information directly from Rocket's website.

Can't understand something? Don't worry!

Confused with technical jargon that no one can explain? Ask the GPT and get answers in a way that you can understand.

Future use case — Find listings on Rocket Homes

I floated the idea to let users find listings on Rocket Homes directly from the GPT. But we made the call to exclude this features from the MVP's scope.

Pitching my Idea to senior leadership!

I had to present my findings and pitch the prototype on a call with senior executives along with the help of my PM. Mind you I was an intern!

RocketGPT is now a live project!!

After getting sign-off from leadership, RocketGPT became a live project in August 2025. Then I worked with my PM to draft the Product Requirements Document and brainstormed on how to assemble a team and the amount of resources needed. Unfortunately, my internship term came to an end and I had to handoff my responsibilities to the team.

Click to read the PRD

Key takeaways

Pitching your idea boldly

There were a lot of moments when I was making the prototype where I wondered if they would take my idea seriously as I was just an intern. But I thought to myself, I got nothing to lose. And it worked out very well!

Numbers don't lie

Every claim I made was backed up with solid research. I want to thank Andrea Chin for helping me access the relevant data and helping me put it together!

< Back to Homepage

Next case study >

NIHAL PALOCAREN