Increasing AI adoption across Rocket
Role
Senior Product Designer
TIMELINE
Jan 2026 to May 2026
Enabling team members to build their own agents
Scope & Contribution
I led the end-to-end redesign of Rocket's internal agent builder (Catalyst), owning all aspects of design and leading research. I worked very closely with product, engineering, research, and an incredible design intern. Our goal was to make creating agents easier so team members could bring automation into their work and spend more time on bigger problems.
Problem
Many team members had worked at Rocket for decades and their tech stack had not evolved much during that time. They were spending too many hours on repetitive and manual work. There was an MVP of Catalyst with the goal of helping team members build agents to automate those tasks, but the platform was complex and required too much technical knowledge. The people who could benefit from automation the most had the hardest time getting started.
Research showed users didn't understand agentic workflows
I worked closely with our research team through multiple rounds of user interviews to unravel the problem. It quickly became clear that users didn’t understand how agents worked or how they could use them to automate hours of work each week.
To make agentic workflows more approachable, I explored reusable templates and suggested prompts. I pushed this further by building a chat-based agent that guided users through the creation process step by step.
"I don't know what type of agent I can create or what I want to automate. I have no idea where to start."
Rocket Team Member
Building a conversational path to creation
Once it had enough context, the chat agent generated a first version with the right tools, triggers, and instructions. I explored a couple versions of the build screen, where team members could review, refine, and finalize their agent. The chat agent remained available in this part of the flow to answer questions and help them iterate.
Turning research into refinements
After I built a prototype of these initial designs (in v0), we did another round of research and usability testing. On the whole, we validated a lot of our design decisions, but it was clear I needed to refine a few things. I added a plan mode to give people a dedicated space to ask questions and explore changes without making them. I added more areas to recommend things like relevant agents, tools, and triggers. I also incorporated familiar AI patterns like contextual prompts to make editing easier.
Building a community of reusable agents
Templates were not easily understood in research, so I decided to evolve them into community agents. This gave team members real and relevant examples they could use or build off of. They could discover agents created by their teammates, search and filter by category, get recommendations based on their role, and easily invite others to collaborate.
"I want to be able to see what people on my team are building to see if I can steal their agents or get ideas."
Rocket Team Member
Making testing easier
The last update before launch was redesigning the testing experience to make it easier for anyone to validate that their agents were working correctly. AI helped by recommending what to test.
k
hours saved
m+
task automated
agents in first 60 days
x
more users
More teams building with AI
The changes had a tangible impact. Team members across the organization are now building agents, saving thousands of hours and automating millions of tasks. More importantly, agents are freeing people up to focus on bigger problems and do their best work.
"Having a guide to walk me through all the steps of building my agent is a game changer. I'm going to save several hours a week."
Rocket Team Member
Increasing AI adoption across Rocket
Role
Senior Product Designer
TIMELINE
Jan 2026 to May 2026
Enabling team members to build their own agents
Scope & Contribution
I led the end-to-end redesign of Rocket's internal agent builder (Catalyst), owning all aspects of design and leading research. I worked very closely with product, engineering, research, and an incredible design intern. Our goal was to make creating agents easier so team members could bring automation into their work and spend more time on bigger problems.
Problem
Many team members had worked at Rocket for decades and their tech stack had not evolved much during that time. They were spending too many hours on repetitive and manual work. There was an MVP of Catalyst with the goal of helping team members build agents to automate those tasks, but the platform was complex and required too much technical knowledge. The people who could benefit from automation the most had the hardest time getting started.
Research showed users didn't understand agentic workflows
I worked closely with our research team through multiple rounds of user interviews to unravel the problem. It quickly became clear that users didn’t understand how agents worked or how they could use them to automate hours of work each week.
To make agentic workflows more approachable, I explored reusable templates and suggested prompts. I pushed this further by building a chat-based agent that guided users through the creation process step by step.
"I don't know what type of agent I can create or what I want to automate. I have no idea where to start."
Rocket Team Member
Building a conversational path to creation
Once it had enough context, the chat agent generated a first version with the right tools, triggers, and instructions. I explored a couple versions of the build screen, where team members could review, refine, and finalize their agent. The chat agent remained available in this part of the flow to answer questions and help them iterate.
Turning research into refinements
After I built a prototype of these initial designs (in v0), we did another round of research and usability testing. On the whole, we validated a lot of our design decisions, but it was clear I needed to refine a few things. I added a plan mode to give people a dedicated space to ask questions and explore changes without making them. I added more areas to recommend things like relevant agents, tools, and triggers. I also incorporated familiar AI patterns like contextual prompts to make editing easier.
Building a community of reusable agents
Templates were not easily understood in research, so I decided to evolve them into community agents. This gave team members real and relevant examples they could use or build off of. They could discover agents created by their teammates, search and filter by category, get recommendations based on their role, and easily invite others to collaborate.
"I want to be able to see what people on my team are building to see if I can steal their agents or get ideas."
Rocket Team Member
Making testing easier
The last update before launch was redesigning the testing experience to make it easier for anyone to validate that their agents were working correctly. AI helped by recommending what to test.
k
hours saved
m+
task automated
agents in first 60 days
x
more users
More teams building with AI
The changes had a tangible impact. Team members across the organization are now building agents, saving thousands of hours and automating millions of tasks. More importantly, agents are freeing people up to focus on bigger problems and do their best work.
"Having a guide to walk me through all the steps of building my agent is a game changer. I'm going to save several hours a week."
Rocket Team Member
Increasing AI adoption across Rocket
Role
Senior Product Designer
TIMELINE
Jan 2026 to May 2026
Enabling team members to build their own agents
Scope & Contribution
I led the end-to-end redesign of Rocket's internal agent builder (Catalyst), owning all aspects of design and leading research. I worked very closely with product, engineering, research, and an incredible design intern. Our goal was to make creating agents easier so team members could bring automation into their work and spend more time on bigger problems.
Problem
Many team members had worked at Rocket for decades and their tech stack had not evolved much during that time. They were spending too many hours on repetitive and manual work. There was an MVP of Catalyst with the goal of helping team members build agents to automate those tasks, but the platform was complex and required too much technical knowledge. The people who could benefit from automation the most had the hardest time getting started.
Research showed users didn't understand agentic workflows
I worked closely with our research team through multiple rounds of user interviews to unravel the problem. It quickly became clear that users didn’t understand how agents worked or how they could use them to automate hours of work each week.
To make agentic workflows more approachable, I explored reusable templates and suggested prompts. I pushed this further by building a chat-based agent that guided users through the creation process step by step.
"I don't know what type of agent I can create or what I want to automate. I have no idea where to start."
Rocket Team Member
Building a conversational path to creation
Once it had enough context, the chat agent generated a first version with the right tools, triggers, and instructions. I explored a couple versions of the build screen, where team members could review, refine, and finalize their agent. The chat agent remained available in this part of the flow to answer questions and help them iterate.
Turning research into refinements
After I built a prototype of these initial designs (in v0), we did another round of research and usability testing. On the whole, we validated a lot of our design decisions, but it was clear I needed to refine a few things. I added a plan mode to give people a dedicated space to ask questions and explore changes without making them. I added more areas to recommend things like relevant agents, tools, and triggers. I also incorporated familiar AI patterns like contextual prompts to make editing easier.
Building a community of reusable agents
Templates were not easily understood in research, so I decided to evolve them into community agents. This gave team members real and relevant examples they could use or build off of. They could discover agents created by their teammates, search and filter by category, get recommendations based on their role, and easily invite others to collaborate.
"I want to be able to see what people on my team are building to see if I can steal their agents or get ideas."
Rocket Team Member
Making testing easier
The last update before launch was redesigning the testing experience to make it easier for anyone to validate that their agents were working correctly. AI helped by recommending what to test.
k
hours saved
m+
task automated
agents in first 60 days
x
more users
More teams building with AI
The changes had a tangible impact. Team members across the organization are now building agents, saving thousands of hours and automating millions of tasks. More importantly, agents are freeing people up to focus on bigger problems and do their best work.
"Having a guide to walk me through all the steps of building my agent is a game changer. I'm going to save several hours a week."
Rocket Team Member
Increasing AI adoption across Rocket
Role
Senior Product Designer
TIMELINE
Jan 2026 to May 2026
Enabling team members to build their own agents
Scope & Contribution
I led the end-to-end redesign of Rocket's internal agent builder (Catalyst), owning all aspects of design and leading research. I worked very closely with product, engineering, research, and an incredible design intern. Our goal was to make creating agents easier so team members could bring automation into their work and spend more time on bigger problems.
Problem
Many team members had worked at Rocket for decades and their tech stack had not evolved much during that time. They were spending too many hours on repetitive and manual work. There was an MVP of Catalyst with the goal of helping team members build agents to automate those tasks, but the platform was complex and required too much technical knowledge. The people who could benefit from automation the most had the hardest time getting started.
Research showed users didn't understand agentic workflows
I worked closely with our research team through multiple rounds of user interviews to unravel the problem. It quickly became clear that users didn’t understand how agents worked or how they could use them to automate hours of work each week.
To make agentic workflows more approachable, I explored reusable templates and suggested prompts. I pushed this further by building a chat-based agent that guided users through the creation process step by step.
"I don't know what type of agent I can create or what I want to automate. I have no idea where to start."
Rocket Team Member
Building a conversational path to creation
Once it had enough context, the chat agent generated a first version with the right tools, triggers, and instructions. I explored a couple versions of the build screen, where team members could review, refine, and finalize their agent. The chat agent remained available in this part of the flow to answer questions and help them iterate.
Turning research into refinements
After I built a prototype of these initial designs (in v0), we did another round of research and usability testing. On the whole, we validated a lot of our design decisions, but it was clear I needed to refine a few things. I added a plan mode to give people a dedicated space to ask questions and explore changes without making them. I added more areas to recommend things like relevant agents, tools, and triggers. I also incorporated familiar AI patterns like contextual prompts to make editing easier.
Building a community of reusable agents
Templates were not easily understood in research, so I decided to evolve them into community agents. This gave team members real and relevant examples they could use or build off of. They could discover agents created by their teammates, search and filter by category, get recommendations based on their role, and easily invite others to collaborate.
"I want to be able to see what people on my team are building to see if I can steal their agents or get ideas."
Rocket Team Member
Making testing easier
The last update before launch was redesigning the testing experience to make it easier for anyone to validate that their agents were working correctly. AI helped by recommending what to test.
k
hours saved
m+
task automated
agents in first 60 days
x
more users
More teams building with AI
The changes had a tangible impact. Team members across the organization are now building agents, saving thousands of hours and automating millions of tasks. More importantly, agents are freeing people up to focus on bigger problems and do their best work.
"Having a guide to walk me through all the steps of building my agent is a game changer. I'm going to save several hours a week."
Rocket Team Member