Noah Larsen Product Designer × Front-End Engineer ← All work
Noah Larsen/Work/Reveal AI

Case Study · 2024–Present

Reveal AI

Helping a services company
become a product company.

Role Product Lead
+ Designer / Developer
Scope Strategy, design,
front-end, AI features
Stack React, LLM orchestration
Engagement Fractional,
direct to CEO

Reveal AI was a services company running qualitative research with hand-built chatbots. To become a product company, they needed customers to design, run, and analyze studies on their own — without anyone at Reveal in the loop.

I joined as the fractional product lead, working directly with the CEO. I set product direction, maintain the roadmap, lead design end-to-end, and ship front-end code alongside the engineering team.

My first initiative was the Interview Designer — the surface that lets a customer compose a chatbot interview from scratch, preview it, and publish it. It’s the feature that unlocked the shift from manual delivery to self-serve. Since then I’ve shipped two AI-powered features in production: a screener that flags likely AI-generated responses, and an analyst surface that turns a study’s transcripts into themed insights and quotes.

01 The Interview Designer is the heart of the product. Researchers compose an interview as an ordered list of question cards, screen out AI-generated participants up top, and lean on the Assistant — a chat sidekick wired to the study itself — to add questions, modify settings, or test the chatbot without leaving the page.
02 Once a study is live, Collect is where it gets monitored. A row of headline metrics — interviews started, included in analysis, median duration, AI detection summary — sits above a sortable transcript table so a researcher can spot drop-offs, terminations, and suspicious responses at a glance.
03 Discover is where the AI analyst earns its keep. The Report view assembles an executive summary and key insights from the full corpus of transcripts — the same artifact a Reveal analyst used to deliver by hand, now generated, editable, and exportable from inside the product.
04 Drilling into a single question surfaces themes and subthemes ranked by frequency, filterable by any other question’s responses. A quotes panel pulls representative verbatims with one click through to context — the move from “we have a thousand transcripts” to “here’s what they said and why.”
05 Chat With Your Data is the open-ended layer on top — ask the corpus a question in plain English and get an answer grounded in the transcripts, with citations. It turns the analyst’s report from a deliverable into a conversation the customer can keep having.
“

From hand-delivered studies to a self-serve product — designed, scoped, and partly built one feature at a time.

— The arc of the engagement

Case study

Photobucket↗

Shipped what two teams couldn’t.

← Back

All work↗

Return home.

Get in touch

Looking for my next
full-time role.

Senior product design and design-engineering roles.

Start a conversation → noah@oomfactory.com