Heizen | AI-Powered Discovery Platform
My Role
Product Designer
Tools Used
Claude · Claude Code · ChatGPT · Vercel
Timeline
1 week
Redesigned Heizen's research workflow into an AI-assisted pre-meeting intelligence system
I redesigned Heizen's research workflow into an AI-assisted pre-meeting intelligence system that transforms public company data into business signals, stakeholder-specific questions, potential gaps, and relevant Heizen proof — helping teams move from “What do we know?” to “What should we ask, and where can we help?”
What is Heizen?
One workflow. From research to opportunity.
Heizen is an internal pre-meeting intelligence platform designed to help teams understand a prospective client before the first conversation.
Instead of manually going through scattered company research, Heizen brings research, business signals, stakeholders, discovery questions, potential gaps, and relevant past work into one connected workflow.
The goal is simple: help teams enter a client conversation with context, not just information.
The Problem
Heizen had information, but lacked a clear path from information to action.
The original experience gave teams access to company research, but preparing for a client conversation still meant navigating large amounts of information, interpreting signals manually, forming questions, and searching for relevant past work.
The challenge wasn't finding more information.It was helping teams understand what matters, what needs to be validated, and where Heizen could potentially help.
The design challenge:
How might we turn scattered company research into a focused, actionable discovery workflow — helping teams understand what matters before they walk into the room?
What I Solved
I restructured the experience around progressive disclosure and contextual decision-making, prioritizing high-signal information first and connecting research directly to stakeholders, questions, gaps, and previous Heizen work. Rather than making users navigate multiple disconnected sections, the redesigned workflow helps them understand what is happening, why it matters, what they need to validate, and how Heizen can potentially help.
Rethinking the discovery workflow
The product concept and initial requirements were already defined. My role was to look at how the experience worked as a whole, identify where the workflow became fragmented, and rethink how users could move from research to meaningful discovery.
Rather than treating research, questions, stakeholders, and previous projects as separate destinations, I restructured the experience around a connected decision-making flow.
What I needed to understand
Each stage answers the next question a team has before a client conversation:
What do we know? → What matters? → Who should we talk to? → What should we ask? → What's missing? → How can we help?
How AI Fits into the Workflow
01 — Understanding the product
ChatGPT · Product & UX Exploration
I used ChatGPT to understand the existing product by sharing its URL and screenshots, then translated my observations into a structured PRD covering the product, workflows, users, and key UX opportunities.
AI didn't teach me the product, I taught the AI what to look for.
I used it as a collaborator to organize my thinking, challenge assumptions, and turn my understanding into a clearer product direction.
02 — Designing & Iterating
Claude Code · From structure to UX refinement
Version 1
I gave Claude Code the PRD, my initial direction, UI references, and requirements. The first version helped me quickly establish the overall product structure and information architecture, but it also exposed where the experience needed stronger UX thinking.
This is where I stepped in. I reviewed the flow myself, questioned the hierarchy, identified friction points, and thought through how the experience should actually work for the user — rather than simply accepting the AI-generated structure.
Version 2 — After UX refinement
I translated those UX decisions back into Claude Code and iterated on the experience, refining the information hierarchy, navigation, interactions, and visual system to create a more purposeful second version.
AI helped me move fast. My role was to decide where it should go.
03 — Micro UX explorations
A) Helix: AI, right where you need it
Helix isn't a separate AI feature hidden away in the product. I made it accessible from the primary navigation so users can quickly ask questions, clarify unfamiliar terms, or explore information without breaking their workflow.
The idea: keep AI available as a lightweight support layer — help when needed, without taking users away from the task.
Micro-interaction: Access Helix → ask → get context → continue working.
B) Keeping research up to date
A quick check before entering the project
After clicking View Project, I added a lightweight check-in for new sources like call recordings, files, or emails. Since research can become outdated quickly, this creates an intentional moment to refresh it, while still letting users skip and continue when there’s nothing new.
The idea: turn research updates into a natural part of the workflow, not another task users have to think about.
C) Making questions feel connected
Questions · From linear lists to contextual discovery
Claude initially explored the questions as a connected tree. I refined this direction because discovery conversations rarely follow a straight line — one answer often determines the next question. Showing the relationship between parent and follow-up questions helps users understand the context and choose where to dig deeper.
I also introduced the info icon to keep additional guidance available without adding noise to every question. On hover, users can quickly understand why a question matters, what to listen for, and what a strong answer might reveal — giving them context without interrupting their flow.
The goal: make the questions feel less like a generated checklist and more like a guided conversation path.
D) Faster navigation, personal comfort
Navigation · Reducing clicks and supporting user preferences
I added a project switcher directly to the top navigation, allowing users to jump between projects without returning to the Projects page — reducing unnecessary navigation and clicks.
I also added a light / dark theme toggle so users can choose the viewing environment that feels most comfortable, especially when working with information-heavy screens for longer periods.
The goal: make frequent actions faster while giving users more control over how they work.
04 — From prototype to product
GitHub + Vercel
After feedback, I refined the final version, pushed the code to GitHub, and deployed the working experience on Vercel.
What I learned
From research tool to decision-making tool
The biggest shift wasn't visual polish. It was changing the role of the product — from a place to collect and browse research into a system that helps teams interpret information and prepare for action.
01 — Structure creates clarity
Connecting related information reduced the cognitive load of moving between research, questions, and opportunities.
02 — AI works best inside a workflow
The value wasn't simply generating content or interfaces; it was using AI to accelerate synthesis, exploration, implementation, and iteration.
03 — Working prototypes change the design process
Building the experience directly with Claude Code made it possible to test interaction decisions earlier instead of waiting until handoff.
Thank you for reading till the end!