Scaling Blue-Collar Recruitment Through Gita AI Interview

Hari Mukti

Completed work

Product Designer

UI Designer

UX Designer

Figma

Jira

Whimsical Wireframes

Human Resources

Overview & Context

The Company | Gapai

Industry: International blue-collar worker placement

Market: Indonesian workers → overseas employment

Scalability becomes a challenge when recruiting in large numbers. With AI interviews, we aim to make the process faster, more accurate, and scalable. Instead of interviewing candidates one by one — which is both time-consuming and exhausting — we designed a system that can handle it at scale.

My role: UI/UX Designer Team: Product Manager, AI/ML Engineer, Backend Developers, HR Stakeholders

🎯 The Challenge

Business: When Growth Becomes a Nightmare

The global demand for Indonesian workers keeps rising — sometimes with requests for more than 600 workers at once.

At the first stage of interviews, we only had two recruiters facing over 1,800 applicants every month. Each interview takes 5–10 minutes, but time was limited. A recruiter could only handle around 40 candidates a day. This bottleneck slowed us down, even when the demand was urgent.

🔍 Research & Discovery

Before jumping into any solution, we needed to understand why hiring was such a mess. So I did what any designer does: I got uncomfortably close to the problem.

Candidates' Interview (n=4)

To understand the candidates’ pain points, I conducted a deep interview with four of them. The goal was to see what actually happens during the first-stage interviews, where the process slows down, and how candidates experience it. Here is the persona.

HR Shadow Session

We are also scheduling interview sessions with our HR team to gain a deeper understanding of how they work (particularly during the first-stage interview process) and to try to feel the challenges and pain points they face.

🤔 Solution Exploration & Decision

With a clear understanding of the problem, we faced a classic startup dilemma: Should we hire more recruiters, optimize what we have, or try something different?

The key questions were:

How might we solve the overload of HR capacity?

How can we provide interview time beyond weekdays and office hours?

Then we try to find possible solutions to solve the pain point and use metrics to decide which solution to implement.

Why AI Interview?

Future-Proofing: Projected growth in demand for Indonesian workers is growing fast

Competitive Advantage: Most of our competitors are still using manual methods

Richer Data: AI interviews provide more comprehensive insights

Candidate Preference: Our internal survey shows that 63% of candidates prefer flexible interview scheduling

The Big Idea

Instead of creating another cold, robotic system, I wanted something that felt more human and supportive — sometimes even warmer than real conversations.

Three Design Principles

AI should help people be better, not replace them

Keep it clear and honest — no tricks, no confusion

Give everyone a fair chance with a simple, accessible design

🎨 Design Process

Conversation Flow Design

We replicated the first interview process into the Gita AI interview with an adjustment, structuring it into 4 main phases:

Welcome screen with brief questions (aimed at reducing the interview agent’s workload)

Tech check (make sure everything works perfectly)

Interview session (consisting of readiness confirmation and the main interview), and finally,

The interview results

Wireframe

The challenge was visualizing an AI that feels like a real HR in static wireframes. I created wireframes as the starting point for discussions on design direction.

✨ Gita AI interview anytime, anywhere

With only two recruiters, we could interview about 80 people per day. But we were getting over 1,800 applications monthly. Absolutely, we needed a better solution.

I designed the Gita AI Interview to feel friendly and helpful, not robotic. It works like the current process, but much faster and available 24/7.

Welcome screen & tech check

The interview starts with a welcome screen that greets the candidate and asks a few opening questions. This helps reduce the technical load and save time in the main session.

An illustration is used to welcome the candidate and confirm their readiness for the interview.

Once all questions are answered, the system automatically checks the internet connection to ensure it’s stable enough to continue or suggests fixing it before moving forward.

Interview Session & the Result

The interview session is divided into two phases. First, the candidate connects with HR 1 to double-check readiness through a short voice conversation.

During the interview, candidates can choose to read text or listen to the interviewer, making it easier for them to respond comfortably.

Once finished, we provide an easy way for candidates to share feedback, as we truly value their input.

Finally, the interview results are shown clearly, indicating the candidate’s level based on well-defined metrics and matched the position.\

🚀 Final Design Solution

Through research and iteration, we built the Gita Interview system with three key features:

👤 Humanized AI Interviewer

📱 Mobile-First Candidate Experience

📊 Automate Dashboard Matching

Automatic matching to job requirements: The HR team doesn’t need to check the candidate's results against the job requirements

🚀 Gita Interview soft Launch & Testing

Versi 1.0 (Current Version)

The focus of this version is on assessing candidates based on their English level and work experience. It will be used alongside the first interview as an alternative option for the first interview phase.

In the soft launch stage, we tested it with our internal team to test the concurrent interview capacity. After that, we moved to external testing with a group of 50 whitelisted candidates, and here is the result.

Versi 1.1 (Next Iteration)

In the next iteration, we will focus on developing planned features and making improvements based on candidate feedback from the Phase 1 testing. These include:

Auto-saving video interviews in the dashboard

Adding animation to the interviewer

Enhancing speech-to-text accuracy

📊 What is the result and the impact?

Quantitative Result

First-stage Interview time from >3 days → 20 Min*

Candidate satisfaction during Interview testing: 4.1/5 average rating

Operating Hours Interview from 8 hours a weekday → Available 24/7

Increase Interview capacity from 40 → Up to 600 per day

Qualitative Result

Better than the last test.... Salute

This latest version might be better. The text of the transcript is good; there are no random translations in other languages, GITA’s communication is also seamless, and her noisy language works very well. Great Job, Product Team!

HR Team

* Conduct AI interviews as soon as the CV screening is successful.

Takeaways 🚴‍♂️

How I Think About AI

This project taught me that the best AI doesn’t try to be human — it tries to simplify human life. Our most successful interviews happened when people forgot they were talking to AI and just focused on telling their story.

The Business Impact That Actually Matters

Yes, we tripled hiring capacity without hiring more HR people. But the real win was creating a system that made everyone’s experience better — for HR, candidates, and the business team.

*Note: For Gita interview testing through Gapai.id/login

Like this project