Designing the First Product at Navless.ai
As a founding team, we transformed Navless from a metaphor to a functional product. By the end of our pilot program, we turned pilot users into paying customers, enabled new sales, and secured $1M in funding.
My role
I worked as both product designer and product strategist. Partnering with the founder, I translated a broad vision into actionable, scalable requirements. Engineering and I iterated quickly to match an evolving sales and investor narrative. By watching visitor sessions and speaking regularly with pilot users, I anchored that pace to a high standard of usability.
Product Marketers ≠ Full Time Content Creators
Storycrafting was a bottleneck for the PMM. They needed a way to tell their product story better by leveraging the content they already had.
Time-strapped
In a survey I found PMMs were under pressure to deliver personalized content, but didn't have the time or bandwidth to keep creating new assets. This was often their #1 pain point.
Bounce-rate anxiety
Through interviews we discovered traditional marketing websites were failing to perform. CMO's were worried by a trend of dwindling traffic and rising bounce rates.
Content-conversion
We found that prospective buyers typically consume ~3 pieces of content before talking to sales, but teams had no way to know which content led to conversions- or how they were connected.
The Netflix of B2B Content
The CEO used the Netflix metaphor as a simple way for stakeholders to grasp the vision. My role was to translate that metaphor into something people could actually interact with.
A visual metaphor
I created an prototype where the UI did a lot of heavy lifting to visually communicate the goals of the experience. This was used to scout for interested pilot users and investors.
Product Pillars for Scaling Our Vision
We knew this first version wasn't the long-term vision. To use it as a meaningful launchpad, four foundational pillars would help us scale the product.
Simple ingestion
Users needed a fast way to import content in bulk.
User-AI alignment
Users and the system needed to be aligned on how to index content.
Smart discovery
The system needed to be informed by website visitor needs and intentions.
Data-driven
Both the system and the user needed the right data to sequence content into a meaningful story.
Simulating Intelligence with Limited Resources
I combined the tools we already had into an experience that felt intelligent. This was a necessary bridge before investing in real ML. A pilot program of this would give us data to reveal where ML investment was actually needed.
Flow state
I designed a multi-prompt flow that ingested content, indexed it, and surfaced it on a marketing site in a way that felt personalized and narrative-driven.
Available technologies
OpenAI API (GPT, Embeddings)
Elasticsearch
Web scraper
v0 for fast UI iteration
Backend
- User defines tagging schema
- User imports content
- AI summarizes each piece
- AI assigns tags based on schema
- User reviews & approves tags
- User installs Navless launching point across website
Frontend
- Visitor launches Navless
- System scrapes the launch page
- AI identifies relevant tags
- System queries indexed content
- Content ranked across three dimensions
- Query grows as user interacts with content
Engagement is More Important than Browsing
Early analytics suggested strong engagement, but a review of 50+ session recordings revealed the clicks weren't intentional. If the first content didn't hook them, visitors defaulted to scrolling or random clicking. Sessions that looked 'busy' weren't usually meaningful.
Redefine performance
We shifted our analytics from click counts to time spent. This favored depth over noise.
Remove redundant affordances
We removed a play button that existed only for tracking and didn't help visitors engage with content.
Introduce pinning
Until the system could automatically feature top content, we gave PMMs manual control through pinning.
Chat is a Path to Engagement
We reframed the search bar as a chat bar as an experiment. Pilot users immediately gravitated toward it, and we saw early demand for "conversational search."
Constraining chat
Pilot users were hesitant about chat misguiding people- or failing to guide them towards proper support channels. To avoid confusion with live support, we limited chat to content-only responses. Quickly, pilot users gained confidence in the chat feature, and requests came in to expand its capabilities.
Embeddable Widgets Encouraged Imagination
This was the key unlock for converting pilot users into paying customer. Embeddable widgets drove the biggest engagement lift and became central to the investor narrative.
Embeddable search bar
Visitors could type directly into a form that launched Navless into 'chat mode' with their query.
Embeddable prompt cards
We prototyped prompt cards inside the Navless interface. Pilot users demanded to embed them on their own sites as direct launching points.
$1M+ in Investments
The product designs became a core backbone of investor pitches. My work iterating on the product led to a clear value proposition.
Real Conversions in Navless
As the product evolved and users became excited to use Navless on their website, they started to see real conversions they could attribute directly to Navless. Some users began planning website changes to make Navless the core experience.
Enabled Rapid Iteration for Engineers Using AI
I owned the core design system, components, animations, and interaction logic which I turned into real code in v0. AI sped up low-level iterations and enabled non-designers to make quick changes on the fly.