Read

As a User Researcher and UX/UI Designer, I worked with government analysts to identify pain points and refine the platform. Through a mix of research methods, I prioritised key issues and guided product development to improve efficiency and trust.

Cyber threat intelligence analysts manually review lengthy reports to extract insights, a slow and overwhelming process. READ aimed to automate this, but analysts lacked confidence in AI-generated annotations and found the platform unintuitive, leading to extra work instead of time saved.

I collected qualitative and quantitative data through workshops and surveys with intelligence analysts to understand their workflow and frustrations. Using thematic analysis, I identified recurring issues and applied evidence-based prioritisation to focus on the most critical UX challenges impacting adoption.

To reduce manual corrections, I introduced Cascading Entities, allowing changes to apply across entire documents. Manual Rules saved annotation preferences, ensuring consistency, while a Performance Dashboard provided transparency on AI accuracy and improvements over time.

These UX improvements removed user frustration, reduced redundant work, and increased confidence in AI annotations. Analysts engaged more with the tool, and measurable improvements in accuracy and efficiency led to greater adoption.

Trust in AI grows with transparency, as users need to understand and refine system decisions. Prioritising usability over pure model improvements unlocked greater efficiency, proving that even the best AI needs an intuitive interface to be effective.