UX Case Study · 2025

Eranova AI

Web page design using Codex

Designing an enterprise-scale AI platform — from blank canvas to a deployed product.

Scroll
www.eranova.ai
Eranova AI
01 — Context

A landing page with
an enterprise-sized opportunity.

Eranova AI had a problem most startups would want: unexpected traction. What started as a simple marketing landing page was pulling in real enterprise clients — but the site didn't look the part. The brand felt generic, the UX was broken in ways nobody had documented, and there was no design system to speak of.

The brief was simple. Make it feel enterprise. Build something that closes deals before a sales call happens.

3
months to audit, rebrand, and launch the full redesigned site
50
branded LinkedIn posts produced after A/B testing content formats
1
master template scaling across all solution and case study pages
02 — UX Audit

Before redesigning anything,
I documented everything wrong.

I ran a systematic evaluation of the existing site using Nielsen's 10 Usability Heuristics — a recognised industry framework for identifying UX failures. Every issue was captured and presented to the CEO in a PowerPoint deck to build alignment before a single pixel was changed.

"You can't redesign something well if you haven't understood exactly why it's broken. The audit wasn't just research — it was the pitch for the redesign."
Nielsen heuristics audit — dark mode Nielsen heuristics audit — light mode

Nielsen's 10 Heuristics — CEO presentation deck, Month 01

H1 — Visibility of System Status

No CTA in the hero

Primary actions — "Get Started" and "Learn More" — were buried at the bottom of the page. Enterprise visitors landed with no immediate next step. First impressions were wasted.

H4 — Consistency & Standards

Broken theme across pages

The homepage was full dark-mode. Inner pages abruptly switched to light mode, only the navbar stayed dark. Jarring and unpolished — the digital equivalent of walking from a cinema into bright sunlight.

H8 — Aesthetic & Minimalist Design

Filler content burying the lead

A full-screen section consumed nearly two scrolls to deliver one generic sentence. It pushed all the valuable content further down and significantly raised the risk of abandonment.

H8 — Aesthetic & Minimalist Design

Weak visual hierarchy on CTAs

"See all solutions" — the key next action for the entire solutions section — was styled as a low-contrast text link. No prominence, no weight. Easy to miss entirely.

Visual Language

Inconsistent icon & font styles

Icon styles mixed freely across the interface. Font weights and sizes had no coherent system, making the site feel unfinished and untrustworthy to enterprise visitors making fast judgements.

Brand Perception

Not enterprise-grade

The overall aesthetic — grey tones, inconsistent spacing, no design system — failed to communicate the quality of the product behind it. The site didn't match the ambition of the business.

03 — Process

Three months.
Three distinct phases.

Each month had a clear focus. No overlapping, no jumping ahead.

Month 01
Audit & Strategy
Nielsen heuristic evaluation, CEO presentation, alignment on redesign direction and brand shift
Month 02
Design & Build
Figma design for all pages, Codex-powered development, master template for all solution pages
Month 03
Brand & Content
LinkedIn A/B testing, 50 branded posts, full brand asset library, colour and type system
Throughout
CEO Check-ins
Structured stakeholder reviews at each phase, feedback loops to keep the work aligned with business goals
04 — Rebrand

From grey and white
to pure black and white.

The original site lived in an ambiguous grey zone — neither confident nor considered. The rebrand made a deliberate choice: strip everything back to black and white. The aesthetic draws from the language of AI itself — binary, precise, zero noise. Clean like code. Trustworthy like a system that works.

Every element either earns its place or disappears. That constraint became the design system.

#0A0A0A
Background
#111111
Surface
#555555
Muted
#888888
Sub
#E8E8E8
Accent
#FFFFFF
White
Before — Dark Grey Charcoal
Eranova before rebrand
After — Pure Black & White
Eranova after rebrand
Before
Grey & white palette, mixed icon styles, no defined type scale, no design system, light-and-dark page inconsistency
After
Pure black & white system, unified icon library, defined typography hierarchy, reusable component library, consistent dark theme across all pages
Concept
The visual language is intentionally binary — drawing from the feel of AI itself. No colour distractions. The product and the content do the talking. Zero and one.
05 — Design & Build

Figma first.
Codex to ship.

The redesign had an aggressive one-month build window. The workflow had to be tight. Every design decision happened in Figma first — layouts, components, spacing, content hierarchy. Only once the design was solid did development begin.

OpenAI's Codex translated the Figma specs into production code at speed. The initial prompts and creative direction were mine. As page complexity grew — especially across the solutions architecture — the senior developer joined to handle advanced implementation while I maintained design ownership throughout.

Key design decision — Rather than designing each solution page individually, I built one master template covering layout, component structure, spacing rules, and content hierarchy. That single template scaled across every product use case without redesigning from scratch each time.

01 Design
Designed all pages in Figma — homepage, solutions index, and the master template — before any code was written. Design thinking always led.
02 Prompt
Fed Figma specs into Codex with structured prompts. Iterated on output to match design intent. AI accelerated execution — it didn't replace design judgment.
03 Collaborate
As complexity increased, the senior developer handled advanced implementation. Figma was the single source of truth. Design direction stayed mine throughout.
06 — Brand & LinkedIn

50 posts.
All data-driven.

The LinkedIn work wasn't production for the sake of production. It started with research. I studied existing post formats and performance patterns for Eranova's audience, then ran A/B tests across different post types before designing a single deliverable. The content strategy came directly from what the data showed — not from assumptions.

01
Research & A/B Testing
Studied existing post performance patterns. Tested different formats — image-heavy vs text-first, short vs long-form, carousel vs static — to understand what resonated with Eranova's specific audience before committing to a design direction.
Complete
02
Content Strategy Redesign
Used A/B findings to define a new content approach — consistent brand voice, visual template system, and a posting cadence that felt intentional rather than ad hoc.
Complete
03
50 Branded Posts + Asset Library
Produced 50 branded LinkedIn posts using the new system, alongside a full brand asset library — colour system, typography guidelines, and reusable components — giving the team a consistent foundation for all future communications.
Complete
LinkedIn · 50 posts. Every one on-brand.
LinkedIn post 1
LinkedIn post 2
LinkedIn post 3
LinkedIn post 4
LinkedIn post 1
LinkedIn post 2
LinkedIn post 3
LinkedIn post 4
LinkedIn post 5
LinkedIn post 6
LinkedIn post 7
LinkedIn post 8
LinkedIn post 5
LinkedIn post 6
LinkedIn post 7
LinkedIn post 8
07 — Design System

Tokens that scale
beyond the project.

One of the most lasting deliverables from this engagement wasn't visible on the website — it was the foundation underneath it. I built out a full token-based design system in Figma: mapped variables for icons, surfaces, and borders, each with defined light and dark mode values. This gave Eranova's internal team a single source of truth they could apply consistently across the site, future product work, and any internal tooling — without needing to re-solve colour decisions from scratch every time.

The system covered 16 mapped variables across three groups — Icon, Surface, and Border — all aliased to a foundation layer so updates cascade automatically. What started as a branding exercise became infrastructure.

Design system variables panel — dark mode Design system variables panel — light mode

Figma Variables panel — Mapped collection, 16 tokens

08 — Reflection

What I carried
out of this.

01 On Audits
Good design work starts before Figma. Building the evidence, presenting it to a CEO, and earning alignment was just as important as anything I designed. Without that foundation, the redesign had no clear direction.
02 On AI + Design
Working with Codex sharpened how I think about handoffs. When your Figma work is precise, AI executes it fast. When it isn't, the gaps show immediately. It made me a more rigorous and intentional designer.
03 On Content
Test before you produce at scale. The A/B data shaped the entire LinkedIn strategy. That discipline — lead with research, follow with execution — is something I'll carry into every project.