Go Discover
SaaS - AI analytics platform
Client
Go 360 (Xerox)
My Role
Contribution
Research
UX Design
UI Design
Design System
Date
Q1 2025
Background
Go Discover is an AI analytics platform where users ask plain-English questions about organisational performance and receive grounded answers from their data.
The initial wireframes were created by the engineering team to prove out the data pipelines and LLM integration. They focused on how the system operates rather than how people would use it day to day. As a result, the ask → answer → refine flow was not front and centre, key actions were easy to miss, and the navigation reflected the build rather than the tasks users came to complete.
Approach
I began with a UX audit to surface the most important issues and set direction. Findings were prioritised from high to low with clear effort and impact notes, then played back to the client to agree the plan. I ran a focused competitor analysis of conversational tools such as ChatGPT, Claude and Gemini to benchmark the ask → answer → refine model,
and adapted the strongest patterns to the Go Discover context.
I then produced high-fidelity screens for all key journeys and linked them into a clickable prototype to validate flow and copy. Once the client was happy with the approach,
I created the design system with tokenised variables ready for production handover in Next.js and Tailwind.
Outcome
A clear chat-style experience that makes asking, understanding and iterating faster, with unambiguous actions and states.
Reduced cognitive load through consistent interaction design patterns, colour roles, simplified control sets and visible clear navigation.
Accessible foundations baked in (contrast, typography, motion/hover behaviour) and a reusable design system with principles that scales to other Go360 products.
UX Audit
A UX review of the platform was conducted using an Heuristic Evaluation. The framework used was ‘Heuristics for Modern Web Application Development’ devised by Andy Budd, it
was used to identify any glaring problems in the interface.
A number of issues were identified that could be addressed to simplify and improve the user experience. These findings gave a clear set of focus areas for the redesign to make the product easier, faster, and more enjoyable to use for all users.
The key findings were:
A general primitive UI design has been used, lacking visual hierarchy, clear layout and defined interaction patterns.
Some design elements (like colour and button placement) create confusion for users and don’t always follow common patterns people expect.
Information can feel cluttered or unclear in places, making it harder for users to read and quickly understand what’s on screen.
Some important navigation and product information is hidden, which may cause frustration or extra effort for new users.
Certain screens use technical language or formatting that slows down reading or makes information harder to process.
There were several accessibility concerns, where some users with visual impairments may find elements difficult to see or use.
Below are some more details of specific areas from the UX audit showing the issues found along side the screenshots of the current version of the platform before the redesign was applied:
Accessibility
Repeated colour contrast failures on interactive elements prevented AA or AAA compliance.
Focus states were inconsistent or missing on key controls.
Small touch targets and tight spacing reduced usability on trackpad and touch.
Keyboard navigation order did not always follow the visual layout.
Error and success states relied on colour alone, with limited iconography or text support.
Use of caps and long sentences
Heavy use of all caps reduced word shape recognition and slowed scanning.
Long, technical sentences increased cognitive load and hid the primary action.
Screen readers can interpret all caps unpredictably, which harms assistive technology support.
Plain English alternatives improved comprehension without losing precision
Navigation
Primary navigation was placed away from core task areas, forcing unnecessary pointer travel.
Account management items were surfaced more clearly than product tasks, which diverted attention.
Labels mirrored internal structures rather than user goals.
Important routes were only available from a single entry point, limiting discoverability.
Interaction patterns
Unusual interaction design is unclear on how to use and causes high cognitive load to user:
Similar-looking controls performed different actions, while different-looking controls performed the same action.
Users were asked to make multiple decisions before seeing any meaningful output.
Validation was late and feedback was subtle, leading to hidden errors.
Destructive actions were not clearly distinguished from neutral or secondary actions.
Progress was easy to lose due to unclear save behaviour and the placement of Clear actions near primary actions.
Design system
The design system was created with tokenised variables and Tailwind alignment for seamless Next.js handover. The following screens outline the tokens, components and usage guidelines.
