Voyance

2017–20

Lead Designer · Techstars S'22 · AI Document Management & Design System

Led design of Voyance Vision — an AI-powered document management system for fintech businesses
Built and iterated Vo, Voyance's design system, from 1.0 through to 2.0
Worked with PMs, data scientists, ML engineers, frontend/backend devs, and the CEO
Context: Nigerian fintech sector, stringent KYC/identity verification requirements

Vision — Product

Trainable AI Models

Businesses could train custom models or use pre-trained models from Voyance's data engineers
Enabled domain-specific document extraction without engineering overhead

Upload, Export & Annotation

Documents uploaded from local or cloud storage; extracted data downloadable in multiple formats
Advanced annotation tools for labelling and improving model accuracy
Added PDF upload support mid-project based on testing — original assumption was image-only

Workflow Integration

Automated data extraction triggered by predefined events in existing business workflows
Reduced manual intervention — system processes documents without human prompting

API Integration

Added after testing revealed businesses needed to embed Vision's AI into their own systems
Enabled programmatic access to document extraction without using the UI

Vision — Design

Research

Interviewed fintech businesses facing KYC and document processing bottlenecks
Key pain points: manual processing too slow and error-prone; high cost of human review; compliance pressure
Two-phase focus: (1) speed and accuracy via AI extraction, (2) automation to reduce manual effort

Process

Created a design strategy document to align vision and principles across teams
Rapid prototyping — tested concepts and iterated quickly on feedback
Collaborated with external data scientists and ML experts during prototype testing
Rapidly upskilled in OCR, machine learning, and AI to design interfaces without knowledge gaps

Vo — Design System

Context

Joined Voyance with no design system — designers made independent decisions on typography, colour, and spacing
Result: inconsistent UX across products; each new component required custom development from scratch
Co-built Vo 1.0 with the Head of Design — shared component library aligned to the engineering codebase

Why Vo 2.0

Vo 1.0 had two critical failures: engineers weren't involved early; marketing alignment was overlooked
Components weren't practical to implement; product and marketing visuals were misaligned
Interviewed engineers, brand designers, PMs, and marketing managers to diagnose root causes
Findings: too many near-identical colour shades, unused components bloating CSS, inconsistent brand across touchpoints

Process

Rebuilt using atomic design — atoms, molecules, organisms, templates, pages
Two audit teams: one reviewed the existing system; one catalogued live patterns across the product
Figma for design libraries, Notion for documentation and cross-team task management

Key Changes

Colour: reduced palette variations from 9 to 5; primary blue WCAG score improved from 4.85 to 7.91 (AAA)
Typography: switched from Proxima Nova to Aeonik Pro — better readability and multi-language support
Buttons: redesigned for consistent states, clearer hierarchy, and better contrast ratios
Grid: 24-column grid for dashboards; 8px spacing scale applied throughout
Iconography: cohesive icon set built on 8px grid with consistent stroke weights

Results