Starting from Zero
We had no existing product, no established user base, no design system, and no historical product data to learn from. As the only designer, I was responsible for shaping the product experience from the ground up, from understanding the problem and defining user needs to designing the interface and collaborating on the MVP.
Instead of jumping straight into Figma, I focused on understanding the problem and our users. I followed the traditional UX process, using AI as a research and brainstorming partner to work faster while keeping design decisions in my hands.
Project Kick-off
My first step was to sit down with the two stakeholders to understand their vision, business goals, and expectations.
Instead of discussing features immediately, I focused on a few important questions.
-
Why interview scheduling, and why now?
-
Who would buy this product, and who would use it daily?
-
What parts of scheduling should AI automate?
-
Where should recruiters remain in control?
-
What would make a company switch from its existing tools?
These conversations helped me understand that we weren't simply building another scheduling tool. We wanted to reduce the coordination work around hiring while making automation predictable, transparent, and easy to override.
That became our initial product direction.
Automate the repetitive work, not the recruiter's judgment.
1. Starting Discovery with Synthetic Users.
One of the biggest challenges was that we didn't have active customers to research. And because Arborloop was a B2B product, I couldn't assume that the person purchasing it would have the same needs as the person using it.
I started by mapping the different stakeholders involved in hiring. I then used ChatGPT and Claude to explore synthetic stakeholder profiles, simulating potential goals, frustrations, and workflows. I explored four perspectives:
To explore their perspectives, I used an AI-assisted approach to build synthetic stakeholder simulations, treating them as hypotheses rather than validated personas.
I wanted to understand their potential responsibilities, goals, frustrations, and where their needs might conflict.
I looked for problems people were already discussing.
I didn't want the synthetic simulation to become our version of reality. Without an existing customer base, I turned to public recruiter and candidate communities. I looked for detailed accounts of actual scheduling breakdowns rather than broad complaints about hiring.
Three publicly available discussions provided concrete starting points:
