"This candidate is clearly a great fit..."

But I couldn't articulate why to the hiring manager fast enough, and I lost the candidate.

Here's the specific situation. I was sourcing a senior server engineer in the gaming industry. On paper, the candidate's experience was slightly short -- the requirement was "7+ years" and they had 5. But when you dug into their actual projects, this person had single-handedly managed a live service with 100,000 concurrent users for 2 years, handling everything from incident response to infrastructure migration. Even among candidates with 7 years of experience, few had this kind of depth.

I knew all of this. But I couldn't deliver a compelling "here's why this person is the right fit" to the client quickly enough. While I was sending resume summaries, explaining over the phone, and putting together additional materials, another search firm submitted their candidate first. That candidate got the interview slot, and my candidate ended up at a different company.

After more than 10 years in the talent acquisition sector, if you asked me to name my most frustrating moments, it would be exactly these. And it wasn't just once.

What Recruiters Actually Do

From the outside, recruiting looks simple. Get a JD, find candidates, make introductions. That's not wrong. But the sheer volume of judgment hidden inside that word "introductions" is much greater than people realize.

Five years at Korea Search, two years at Jobplanet, then global firms, entertainment, fintech. Across every role, the core of my daily work was bridging three gaps.

First, the gap between what the JD says and what the client actually wants.

When you actually talk to hiring managers, the conversation diverges from the JD. In some cases, the most important criteria isn't even mentioned in the JD at all. Uncovering the specific expectations hidden behind vague phrases like "culture fit" -- that's the recruiter's job.

Second, the gap between what's on the resume and actual competency. A resume isn't a self-introduction letter. Especially in Korea, developers consistently understate their accomplishments. Someone who writes one line -- "Responsible for server development" -- may have actually handled everything from architecture design to deployment automation. Identifying this before the interview and presenting to the client with evidence that "this person is much stronger than their resume suggests" -- that takes significant time.

Third, the strengths candidates fail to articulate themselves. This is the hardest gap to bridge. But also the most rewarding when you get it right. Sometimes candidates don't even recognize their own strengths. Especially those who've moved into management and think they "only did management" -- when they actually have experience building organizations from zero to one, including hiring strategy.

Analyzing these three gaps mentally, organizing them manually, and explaining them to clients or hiring managers -- this process consumes more than half of a recruiter's working time.

What I Learned at Jobplanet

From 2018 to 2020, I worked as Consulting Director at Jobplanet (Korea's equivalent of Glassdoor + recruiting platform) and hit #1 in KPIs two years running. I'm not bragging -- I'm sharing because that experience became the starting point for what I'm building now.

The reason I hit #1 was simple. I analyzed faster and more accurately than anyone else. I had my own framework for reviewing a candidate profile against a JD and organizing strengths, weaknesses, and risks within 30 minutes. It was paper-and-Excel, copy-paste-into-email style, but the judgment structure itself was fairly systematic.

The problem was, this framework existed only in my head. When I tried to teach it to team members, the "intuition" component was too large. Every time I tried to explain "why this candidate is good," I had to unpack my experience-based judgment from scratch, and information was inevitably lost in the process.

That's when I first thought: What if I could turn this judgment structure into a system?

Of course, it was just a passing thought at the time. I had a CS degree and started my career in development, but my hands-on engineering experience was brief and from my junior years. I'd been in recruiting for over a decade. Building a product by myself was unimaginable.

A Market Where Speed Is Everything

Entertainment, gaming, fintech. Across every industry I worked in, one pattern held true. The senior tech talent market is a speed game.

When a strong candidate hits the market, the decision is made within a week. As a headhunter, you're reviewing dozens of candidate profiles per day, manually analyzing each one against the JD, writing individual recommendation comments, and preparing briefing materials for hiring managers. Even at top speed, this takes 30 minutes to an hour per candidate.

In those 30 minutes, another agency submits the same candidate first. Even if I have a more accurate analysis, being late means being irrelevant. During peak hiring seasons, this pressure becomes extreme. You're reviewing 3-5 candidates for a single position simultaneously, and you need to articulate "why this person" persuasively for each one -- but there simply isn't enough time.

After direct-sourcing 15 to 27 tech hires per year, the feeling was always the same. "The time spent analyzing and explaining is too long." And this time doesn't meaningfully decrease with experience, because every JD, every candidate, every context is new.

But the Market Is Frozen Right Now

There's something I need to address here.

I just called it a "speed game," but honestly, the 2025-2026 hiring market has seen that speed game itself shrink significantly. According to the Korea Employers Federation, only 60.8% of companies plan to hire new employees -- the lowest since 2022 -- and 37.8% plan to reduce headcount. Globally it's no different: over 5,000 companies announced mass layoffs in 2025 alone.

The developer market is even more pronounced. According to the Korea Labor Institute, the share of entry-level hires in software development dropped from 53% in 2022 to 37% in 2024, with 9,000 fewer junior hires year-over-year. Meanwhile, experienced hires with 3+ years grew by 42,000 in the same period. As demand concentrates on experienced professionals who can leverage AI, the market itself is being restructured.

"So is this tool even needed right now?" is a fair question.

I believe it's needed now more than ever.

Companies' hiring criteria have fundamentally shifted. The first question isn't "who do we need for growth?" but rather "can this position be replaced by AI or automation?" According to Resume.org, 45% of hiring managers say "AI has reduced the need for new hires," and 30% of global executives are lowering hiring rates because of AI.

The positions that survive are heavy. Average time-to-fill has stretched from 44 days to 63-68 days, with 60% of companies reporting "hiring takes longer than last year." Interview rounds have increased, verification steps have been added, and more stakeholders are involved in decision-making. Cost-per-hire for technical roles averages $6,200, and a failed hire costs 30-150% of annual salary according to the U.S. Department of Labor.

The nature of requests coming to me has changed too. It used to be "recommend someone good if you have anyone." Now it's "give us the justification to convince leadership why this person is needed right now." The recruitment report has gone from being a profile summary to essentially a hiring justification document.

For headhunters, the picture is the same. The number of available mandates has shrunk while the number of competing agencies hasn't. In this market, "accurately explaining why this person is the right fit" matters far more than "submitting fast." Not submitting 5 candidates and placing 1, but submitting 2 and placing 1. That accuracy comes from depth of analysis.

What Existing Tools Can't Solve

There's no shortage of AI recruiting tools. LinkedIn Recruiter, HireEZ, Gem, SeekOut. There are dozens globally, and countless more including regional solutions.

But most of these tools focus on "finding candidates." They're essentially smarter boolean search engines. In practice, however, the real time sink isn't finding candidates -- it's "explaining why this candidate is the right fit."

Having used many of these tools, I kept hitting the same walls.

The keyword matching trap. Search for "Java 7 years" and you get people who've used Java for 7 years. But what the client actually needs is "experience with JVM tuning in high-traffic environments." Keywords and competencies are different things.

The limits of scoring. An 80 isn't necessarily worse than a 90. If Candidate A scores a consistent 80 across all criteria, and Candidate B scores 100 on two critical areas but 60 on the rest, in practice Candidate B is often more compelling. Numbers alone can't capture this judgment.

The absence of context. Five years at a startup and five years at a large corporation are entirely different experiences. Someone who did full-stack work on a 5-person team and someone who owned a specific module in a 500-person organization have completely different capabilities under the same "5-year developer" title. Reading this context comes from a recruiter's experience.

Ultimately, I concluded that the problem is that the people building these tools aren't recruiters. The tools are technically sophisticated, but they don't reflect the judgments recruiters actually make in practice. I personally believe the strongest usability comes when the tool is built by its own power user.

So I Decided to Build It Myself

It wasn't "I'll learn AI coding and build a product." The starting point was "let me try to systematize the TA methodology I've built over the years."

Honestly, I was scared. Through tech recruiting I'd been keeping up with technology stacks and architecture trends, but knowing about something and building it yourself are entirely different. The person who used to ask candidates "do you have experience with this framework?" now had to use that framework to build a product. The cognitive dissonance was real.

But AI in 2025 was a completely different world from when I'd been an engineer. When I described the functionality I wanted to an AI, code came out. At first I thought "is this actually working?" but within a day I had a working prototype that could take a JD and resume and produce an analysis. That's when I saw the possibility.

The key was that I didn't write the code -- I designed the methodology. "What to extract from a JD," "what to look for in a candidate profile," "what structure makes a matching analysis actually useful to a recruiter." These were areas AI couldn't handle alone. You need the sensibility of someone who's done thousands of matches in the field.

AI wrote the code. I designed the analysis framework. This combination worked surprisingly well.

It Won't Be Easy, But

As I write this, Candidate Analyzer is still in its early stages. The core feature is straightforward: input a JD and resume, and it produces a matching report through a 3-stage AI analysis pipeline.

But this "3-stage" structure itself was born from 12 years of experience.

Why three stages instead of one? Because when you ask AI to do it all at once, it lies.

This phenomenon is called "hallucination," and I'll explain the structural approach to preventing it in the next post.

It won't be an easy road. But seeing the things that have lived only in my head for 12 years finally take shape in a working system -- that has given me conviction. The conviction that this is worth pursuing.

I plan to document this journey. There will be more things that don't work than things that do, but I believe the process itself will be useful to others facing the same challenges.