Case study 02 · AI Assistant

Rose

An AI sales agent that embeds on any B2B website and turns visitors into qualified conversations.

Role
Founding Designer
Timeline
Since 2025
Scope
Product · UX/UI · Design System
Platform
Widget · SaaS
My contribution

I design the product from the ground up.

As Founding Designer, I shape Rose end to end, across both the customer-facing agent and the internal platform teams work in.

What I do
  • Product strategy
  • UX & interaction design
  • Visual design
  • Design system
  • Embedded AI experience
  • Analytics platform
The challenge · Designing AI that feels native

One product, two very different audiences.

Visitors reach Rose straight from a company's marketing website, looking for quick answers and guidance. Marketing and growth teams use the platform to understand those conversations, qualify leads and learn from visitor behaviour. The challenge is designing one product that feels native on every client website, while turning conversations into insights teams can act on.

Visitors

People visiting a company's website. They need fast, relevant answers without feeling like they're talking to a traditional chatbot.

Marketing & growth teams

Teams responsible for acquisition and conversion. They want qualified leads, but also a clearer picture of what visitors actually ask before becoming customers.

Design principles

Three principles behind every decision.

Native by default

Rose never feels like a third-party widget. It looks like it belongs to the website where it lives.

Conversations before forms

The experience invites a natural conversation instead of pushing visitors through a static funnel.

Every conversation creates value

Each interaction produces useful knowledge for the team behind the site, not only a potential lead.

01 · The Smart Bar

A sales agent that feels native to every website.

A floating chatbot in the corner reads as a third-party tool. So I make the opposite bet: Rose opens as a Smart Bar in the centre of the page, the way a search field invites you in. Two tensions shape the design. Catch attention without becoming intrusive. Adopt each client's brand without losing one coherent interaction model.

White-label. Rose adopts each host site’s primary colour. Here, it wears this portfolio’s.

Native branding

Rose adapts to each client's visual identity, so the experience feels integrated rather than embedded.

Progressive layout

The conversation begins centred to capture attention, then moves to the side to coexist with browsing.

Guided navigation

Responses can include contextual links that help visitors reach the right page naturally.

Lead qualification

The agent answers questions while spotting qualified prospects and nudging the right visitors to book a demo.

02 · The platform

Every conversation becomes a learning.

Rose doesn't stop when a conversation ends. Every interaction feeds an analytics platform I design for marketing and growth teams. Instead of leading with conversion counts, it surfaces what visitors actually ask, what they struggle to understand, and where the website fails to communicate its value. The job is to turn qualitative conversations into insights teams can keep improving from.

From metrics to meaning

The dashboard opens on the questions visitors ask most, not on a wall of conversion counts.

Themes, not transcripts

Conversations cluster into recurring topics and objections, so a team sees the pattern without reading every chat.

Traceable by design

Each theme links back to the real conversations behind it, so any finding can be checked at the source.

03 · Self-serve

Onboarding a site in minutes.

A company deploys Rose without help. The onboarding guides them from analysing their website to configuring the branding, then generates the final integration snippet. The goal is to make it feel as simple as adding any modern SaaS integration.

Impact
0%
of visitors convert through Rose

Rose turns a passive marketing website into a conversation that qualifies visitors and moves them toward the next step.

Visit userose.ai ↗
PennylaneLive
PayFitLive
SkelloLive
What I learned

The best AI experiences feel the least intrusive.

Designing AI assistants isn't only about conversations. It's about fitting into an existing product ecosystem while creating value for two audiences at once, blending into the workflows people already have instead of asking them to learn something new.

Next projectKolecto