Foundational design for AI startup

Yareta is a seed-funded startup providing predictive AI insights on candidates and employees. Users upload everyday work documents, meeting transcripts, videos, presentations, CVs etc, and get AI-powered behavioural insights benchmarked against the company's custom benchmarks.

The insights are powerful but their UI was confusing clients. I worked with Yareta's CEO, Product Success Manager, and Founding Psychologist to completely redesign the UX in just 8 working days.

The original UI had been designed and developed by a non-designer using Lovable. The UX was characterised by unnecessary complexity, clutter and fragmented flows. The problems had become obvious during recent client trials, and Yareta realised designer input was essential before the next round of client trials.

I began by analysing the existing builds, understanding the functionality, and cataloging the UX issues. Key problems included:

• Overwhelmingly busy UIs with weak hierarchy — rarely clear where the user should be focusing.
• Confusing and misleading hierarchy — e.g. hugely emphasising "meeting date" on a profile page that isn't tied to any one particular meeting — compounded by unnecessary use of ambiguous feature names (e.g. "Success Profile", "Success Builder")
• Maze-like navigation structure — user can't get an overview of what's available.
• Weak and unattractive brand — no strong personality coming through.
• Confusing copy — e.g. "ongoing" as a section title for an overall assessment of a given person, not tied to particular document.
• No reassurance about privacy and data access, a key concern for this sector.

Moving on to solutions, I began by agreeing with the CEO who the core target user was and what the "red route" — the optimal user flow — should look like at a high level. From there I was able to propose a complete reworking of the app structure.

Due to the specific needs of a major tech corp that Yareta was pitching to, the red route centred around a head of HR struggling to hire candidates who matched with the company's specific culture. The user needed to quickly experience candidate analysis, and to be guided through creating company-specific assessment benchmarks.

Using the existing barebones marketing brand, and a shadcn/ui-based Figma library (because Lovable is trained principally on shadcn/ui), I created a basic design system and got stuck into designing the red route UI.

My focus was on reducing the number of clicks to the "wow" moment, creating a clear navigation structure, and developing an attractive look-and-feel, with each page embodying a decisive and informative hierarchy.

The first challenge was separating the genuinely useful content from the noise — the origin designs buried real functionality under layers of visual and structural bloat. Once I'd identified what actually mattered, I could reorganise the content and flows around real user needs. More often than not, that meant cutting. What looked like a feature-rich interface was largely unnecessary bloat.

The design went through several rounds of iterations, first mainly from Yareta's in-house psychologist, and her experience of client needs, and then based on initial client feedback. The biggest change I made was to replace the multi-step onboarding wizard with empty states, after I realised that the initial onboarding was an unnecessary complication.