AI in recruitment: 159 applications filtered down to 1 hire

Quick summary: AI in recruiting can now handle the whole front end: running ads to attract candidates, screening applications, scenario-based trial interviews, scoring, emails and scheduling. Real numbers from MONA: 5,000,000 VND in ads, 159 applications, AI filtered down to 43, and 1 hire.

When it comes to AI in recruiting, there are two extremes online. One side shouts that AI will replace HR entirely. The other says AI screening just wrongly rejects talented people. Both speak from gut feeling, while McKinsey's State of AI survey ranks recruiting among the business functions adopting generative AI fastest. In this article we speak from a real campaign MONA just ran for itself, with a video of Hùng, MONA's founder, opening his screen and walking through each number. Watch 5 minutes and draw your own conclusion.

Give AI the most tedious step, not the whole process

The first principle, also the line Hùng closes the video with: recruiting automation should solve the most tedious, labor-intensive and error-prone part, which is initial screening. The in-depth interview with the manager is still a real person sitting with a real person.

Why is initial screening the most error-prone? Because its volume is inhuman. Sales, customer care and service roles can receive hundreds to thousands of CVs a week. No one can read that many closely, so everyone skims, and skimming both misses good people and invites the wrong ones. Hand exactly that chunk to the machine, and people are free to do what machines can't: spot potential and make decisions.

Inside a campaign: ads that learn on their own, and 4 layers of screening

A complete AI recruiting campaign runs like this. The input is ads on Facebook, Google, YouTube or TikTok. What differs from ordinary recruiting ads: the system AI recruitment software hooks directly into the ad platform, and for every qualified candidate the data is sent back so the ad engine learns. The longer it runs, the better it targets the right audience and the cheaper each candidate gets.

Candidates who click the ad go straight into the assessment, no waiting for an appointment. After verifying with a Google account or an OTP code, each person goes through 4 layers of evaluation: the machine matches the CV against the criteria; a scenario task where the AI plays a difficult customer who gets angrier or calmer depending on the quality of the answers; a psychology quiz designed with no obvious answers to measure the real person; and finally a few questions on motivation and long-term commitment to the job. The approach of “let the machine screen, let people decide” is also what the series from Harvard Business Review on recruiting has recommended for years.

The machine scores everything out of 100 on the server, and candidates can't see their scores, so there's no gaming it. Each application comes with red flags if any and a 3 to 5 sentence summary. We break down the AI interview mechanism separately in a recruitment AI Agent, and the background on resume screening systems is in What is an ATS. Two other deep dives: What is AI recruitment and automated AI interviews.

159 applications, 1 hire, and why that is a good number

The real numbers from MONA's one-week campaign recruiting a sales employee: total ad spend 5,000,000 VND. Facebook Ads reported about 12,000 VND per candidate click into the assessment. 159 people completed it, and the machine scored 43 above the threshold. The manager reviewed by hand and invited 27, 15 booked interviews themselves, and 1 person accepted the job.

At first glance, 159 down to 1 looks wasteful. Look again and it's a good number: none of those 159 applications cost HR a single hour of manual reading. 5,000,000 VND for one hire, compared with recruiter fees that typically run 1 to 2 months of salary, is a huge gap. And the machine doesn't switch off: ads keep running, and once the team is full you raise the score threshold to take only stronger people, or lower it when you're short on staff. Turn one number and you tighten or loosen the whole funnel.

The fear that “AI wrongly rejects talented people” is handled by design, not by promises

The straight answer: sloppy AI screening is real with old keyword-based filters, and the fix isn't to trust the ads but to look at the system's design. A decent system must have four safety valves.

One, applications below the threshold but with a bright spot, for example a weak assessment but a strong CV, must be kept for a person to review by hand. Two, when the machine scores wrongly there must be a re-score button. Three, the real interview is recorded, transcribed and independently scored once more, so supervisors can review it instead of relying on one person's gut. Four, the final decision always belongs to a human; the machine only filters the rough part. MONA tells its own customers plainly: letting AI make the final call is the fastest way to miss talent, so don't do it.

As for candidates, what damages an employer brand isn't a rejection but silence. The system sends step-by-step emails automatically, including a polite rejection timed 60 to 120 minutes later. Every candidate gets a proper reply, so hiring leaves no bad word online.

Which industries can apply it right away, and how much to start

This applies directly to every industry that hires in volume and continuously: retail chains, restaurants, spas and aesthetics clinics, distribution companies that need a telesales team, training centers hiring teachers by season, factories hiring workers by order. For software spending in Vietnam, see industry data on Statista. Swap the criteria set and the scenario bank by industry and it runs, on the same foundation as the other agents in MONA AI Agent suite, from chatbot AI thu lead to automated quotes.

On cost, you can start with just one urgent opening: you set the ad budget, 5,000,000 VND a week as in the campaign above is a reasonable reference, and the system is priced by deployment package. What matters is measuring every dong: how much per application, per qualified candidate, per hire. If you want to go further, read more on AI transformation roadmap for businesses or the full HRM picture.

Frequently asked questions about AI in recruiting

What does AI in recruiting actually do?

It covers running and optimizing recruiting ads, screening CVs by criteria, scenario-based trial interviews, assessment quizzes, candidate scoring, step-by-step emails, interview scheduling, and recording and transcribing interviews for round-two scoring.

Can AI replace the interviewer?

No. AI replaces the repetitive screening and paperwork, about 80% of the volume. The decisive interview is still a manager sitting with the candidate, and the machine only helps record, transcribe and summarize for supervisors to review.

How much does AI recruiting cost?

Real numbers from a MONA campaign: Facebook Ads reported about 12,000 VND per candidate click, and a total of 5,000,000 VND in ads until one person accepted the job. The figure varies by industry and role, but it is always clearly measurable.

Can a small business use it?

Yes, as long as you hire continuously. The system is priced by package, with no infrastructure investment to start. Call 1900 636 648 or book a free consultation, and we'll open a live dashboard to show you before you decide.

See it with your own eyes, then believe it

AI in recruiting is no longer a future story, it's hiring for MONA every day. A better way to verify than any article: watch the video above, then visit the Gấu Tuyển Dụng page to see each feature, or tell us the role you're hiring for and we'll show you the dashboard with real numbers. Call 1900 636 648, answered during business hours.