Designing Cord’s Market Intelligence tool meant translating complex recruitment data into a seamless, conversational interface - similar to ChatGPT - that recruiters could trust for fast, actionable insights. My approach focused on balancing usability, trust, and clarity while working closely with engineers, product managers, and end users.

I began with user research: speaking directly with recruiters to understand how they currently gather market insights like salary data, role demand, and competitor benchmarks. The common thread was inefficiency - users were jumping between tools, reports, and Google searches. There was also a high cognitive load when trying to interpret raw data.

Armed with these insights, I mapped out key use cases and user intents, then worked on crafting natural language interaction flows that felt approachable but remained focused. I created a series of wireframes and prototypes to explore different interface models, ranging from a full conversational UI to a hybrid input with structured suggestions and filters. Ultimately, I designed a clean, prompt-based interface that allowed for free-form questions but offered smart autocomplete suggestions to guide users and reduce errors.

I also worked closely with engineers to shape how responses were presented. Raw text outputs felt too opaque, so I designed modular response components - tables, charts, and concise summaries - so that data was visually scannable, credible, and easy to act on.

Problems Faced & How I Solved Them

1. Ambiguity in User Queries
Users often asked vague or broad questions (e.g., “What should I pay a software engineer?”). This made responses inconsistent and less useful.

Solution: I designed contextual prompts and suggested queries to guide users toward more precise inputs. I also included lightweight follow-up questions from the AI to clarify ambiguous queries.

2. Trust in AI-Generated Data
Recruiters were hesitant to trust AI-generated answers without knowing the source.

Solution: I added transparency features, like “data sourced from…” tooltips, and made it clear when data was based on real-time insights vs. historical averages. I also used UI treatments (color, tone, spacing) to separate facts from suggestions.

3. Speed vs. Usefulness
There was a tension between delivering fast responses and delivering good ones.

Solution: I introduced skeleton loaders and subtle conversational delays that gave the system enough time to generate high-quality answers, while still feeling responsive and fluid to the user.

This project challenged me to design not just a product interface but a conversation model, ensuring that the AI felt helpful, transparent, and trustworthy in a space where accuracy is critical. By staying close to real user needs and iterating quickly with the team, we built a tool that meaningfully improves how recruiters make market-driven decisions.