All projects

Case study 01 · Arborloop AI-native · 0→1

Designing an AI-native interview scheduler from 0→1

Turned an early stage startup idea into a user-centred product experience, using AI as a research and design partner throughout the process.

The manual scheduling problem: a team calendar full of conflicts, an email thread chasing availability, and an interview tracker
Role
Founding Product Designer
Team
1 Designer, 2 Founding Stakeholders, 2 Engineers
Scope
0→1: no product, users or design system yet
Industry
HR Tech · Recruiting
Project overview

Starting from zero

Role

Founding Product Designer

Owned end-to-end product design, from strategy to user flows, interfaces, and a scalable design system.

Team

1 Designer + 2 Founders + 2 Engineers

Worked directly with the founders and engineers to turn an early stage idea into a clear product vision and MVP direction.

Starting point

No product, no users, no system

I established the design foundations, explored user needs, and shaped the experience from scratch.

Approach

AI-assisted design from day one

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 shaped 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.

Discover → Define → Ideate → Design → Validate → Deliver, with AI in action at each step
Getting started

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:

  1. Why interview scheduling, and why now?
  2. Who would buy this product, and who would use it daily?
  3. What parts of scheduling should AI automate?
  4. Where should recruiters remain in control?
  5. What would make a company switch from its existing tools?
Brainstorming with the founders
Three-person card sort
The discovery workshop board

These conversations helped us define our direction: reduce manual coordination while keeping AI transparent, predictable, and under the recruiter’s control.

Our guiding principle

Automate the repetitive work, not the recruiter’s judgment.

Discovery · 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, then used ChatGPT and Claude to explore synthetic stakeholder profiles, simulating potential goals, frustrations, and workflows across four perspectives:

I treated these AI-assisted stakeholder simulations as hypotheses rather than validated personas. I wanted to understand their potential responsibilities, goals and frustrations, and where their needs might conflict.

The key tension it surfaced: automation vs. control

“How might we automate scheduling without making recruiters feel that they’ve lost control?”

Discovery · 2

Looking 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, looking for detailed accounts of actual scheduling breakdowns rather than broad complaints about hiring.

AI-assisted desk research prompt

Three publicly available discussions provided concrete starting points:

Each card opens the original Reddit thread. Individual posts are signals, not representative user research.

Discovery · 3

Public complaints were useful. I also needed scale.

To balance individual stories with quantitative evidence, I studied Ashby’s Recruiting Coordination report, published on 22 October 2024. It analyses scheduling for more than 2.8 million candidates and 5.1 million interview events.

The opportunity

Rather than simply automating interview bookings, Arborloop could simplify complex scheduling workflows, enable candidate self-service, and give recruiters visibility and control over AI-generated schedules.

The bars above are redrawn from published figures rather than screenshots. View the original Ashby report, charts, definitions and methodology ↗

Discovery · 4

Competitive analysis

I used AI to speed up the first pass of competitive research, then verified claims against the companies’ own product and help documentation. I focused on ModernLoop, GoodTime and Calendly because they represent different ways to solve scheduling, from interview orchestration to flexible booking.

ModernLoop scheduling interface
ModernLoop
GoodTime scheduling interface
GoodTime
Features vary by plan and configuration
Competitor research prompt
Synthesis

Discovery takeaway: what I learned

I consolidated my research findings, insights and supporting resources into a single document, then used ChatGPT to analyse and distil them into a handful of actionable insights.

ChatGPT · Interview prep synthesis
  1. Reduce manual work: simplify interview coordination and scheduling.
  2. Handle complex scheduling: make panel coordination and availability matching easier.
  3. Keep recruiters in control: let them review and override AI decisions.
  4. Improve candidate experience: enable easy scheduling, rescheduling and communication.
  5. Build trust in AI: make automation transparent and predictable.

I used these insights to guide my next steps, define the core problem, and prioritise features for the MVP.

See where it landed

Explore the clickable prototype, or try the live product.

The Arborloop recruiter dashboard on a laptop

What’s next?

Let’s design something meaningful.