Most AI interviews give you a number. This one shows you where it came from.
A structured first-round interview, recorded and transcribed, with questions tied to the skills in your competency framework. Every competency judgment on the scorecard is pinned to a second of the recording and the sentence that produced it. Open it and you’re there.
- Tanca HRM
- Skillify
- Vingroup Retail
- Masan Consumer
- GrabFood VN
- Golden Gate
Forty first-round calls, and notes nobody can compare.
The first round is thirty minutes of the same six questions, run by whoever is free, written up from memory afterwards. Two candidates get asked different things, judged against different standards, and summarised in two lines each.
Then a month later someone asks why the third candidate didn’t go forward, and the honest answer is that nobody remembers.
This is one of the three features with an allowance.
Most of Career GPS is unlimited. This isn’t, and we’d rather you knew that on the feature page than discovered it in month two.
Draws on your AI allowance
Unlimited, permanently
Running out means new interviews pause. It never means losing what you already have.
Structured, but it listens to the answer.
Six questions from your framework, plus follow-ups the system decides on based on what the person just said.
Before it starts
Everything that matters is on one screen, in plain words, before the first question — including the consent to be recorded and the number of days the recording is kept.
During
One question on screen, a progress marker, and a visible recording indicator. There is no countdown on individual answers.
Describe a distributed system you’ve designed, and the hardest problem you had to solve.
The follow-up
The second question wasn’t scripted. It was generated from the first answer — and it’s what produced the level-3 judgment on system design.
Open the number. Land on the sentence.
This is the whole difference between a score you use and a score you have to trust.
When a candidate asks why.
You have the sentence, in their own words, at a timestamp. Not a recollection and not a number.
When your hiring manager disagrees.
They can watch ninety seconds instead of taking your word for it, and the argument becomes about the evidence rather than about who is more senior.
When you need to check the system itself.
Read the pins across twenty candidates and you’ll see quickly whether the scoring is sensible. A number you can’t inspect is a number you can’t audit.
AI is a decision-support tool. The analysis below is for reference — AI does not decline anyone automatically. The final decision is yours.
“I’ve mostly worked alongside a team lead; I haven’t led a full system design on my own”
— gap against level 4“I’ve used Express mainly; I haven’t built anything with NestJS yet”
Scored against the framework you wrote.
The questions come from the skills in your competency framework, and the results come back as a level per skill against the level that role requires. L3 / required L4 is actionable. 72% is not.
It’s the same five-level scale as your assessments and your performance reviews, which means an interview result and a six-month review are directly comparable.
The overall match score is a summary of those levels, not a separate judgment. Both are shown, and the page always says which is which.
The AI recommends. It has never declined anyone.
Five things you can do with a completed interview. All five are triggered by a person, and the product says so on every scorecard rather than only in a policy document.
To offer
Move straight to offer, with the recording attached to the record for whoever needs it later.
To a human panel
The most common outcome. Your interviewers arrive already knowing which two competencies to probe.
Interview again
Connection dropped, the person was clearly unwell, a question misfired. Re-run it at no cost to them.
To talent pool
Good, wrong role, wrong time. Kept with consent and surfaced when something fits.
Not continuing
A person decides this, records it, and the candidate is told in plain words the same day.
What the system will never do.
It will never decline a candidate, never hide one from your pipeline, never rank your candidates against each other in anything the candidate sees, and never act on an integrity signal by itself. The scorecard carries a recommendation because a recommendation is useful. It has no ability to execute one.
Indicators for a person to check. Never a verdict.
Four signals come back with every interview: whether the person matched their verified profile, whether they stayed in the interview window, whether one voice was present, and an estimate of how likely the answers were being read from a script.
The fourth one is a probability and it’s shown as a probability — low · 12%, not “clean” and not “cheating”. A number with uncertainty attached is honest. A verdict is not, and a verdict is what gets a company into trouble.
None of them decline anyone. If something looks off, the recording is right there and a person watches ninety seconds. That’s the entire workflow.
Integrity signals are indicators for a human to review. They are never a reason to decline a candidate on their own.
Watch 08:10–09:40You are recording a person’s face and voice. Here’s how that’s handled.
Under Vietnam’s Personal Data Protection Law, biometric data is sensitive personal data. It requires its own explicit consent — not a line inside a general application form — and it requires a retention limit you can state.
Separate, explicit consent for the recording.
Asked on its own screen before the interview begins, in plain language, naming what is recorded, why, who can see it and for how long. Declining doesn’t remove the candidate from your process — it routes them to a human interview instead.
A retention limit, stated up front.
Recordings delete automatically after 180 days by default, configurable by you. The candidate is told the number before they start, and it appears on the scorecard next to the score rather than in a settings page.
A named audience, not a general one.
Only the roles you’ve granted access — HR and the hiring manager by default — can open a recording. Every view is logged.
Deletion on request, at any time.
From you or from the candidate. Early deletion requires a recorded reason, so the log stays meaningful.
What we don’t do.
We don’t score emotion, personality, confidence, enthusiasm, or anything described as cultural fit. Those are proxies for age, background and health, and there is no version of them we’d be comfortable defending. The interview scores the competencies in your framework, and nothing else.
This describes how the product works and is not legal advice. Our consent flows are reviewed by Vietnamese counsel.
The retake nobody can buy.
The common business model in AI interviewing is to give the software to employers free and sell the candidates small things: the right to redo a question, to re-record an answer, to complete a question the timer cut off.
It sounds harmless until you look at what it does to your data. The candidates who paid submitted a fuller, more polished answer set than the ones who didn’t. Your scores are no longer comparable, and the difference tracks who could afford it rather than who can do the job.
Career GPS does not sell candidates anything. Not a retake, not a do-over, not a look at their own results. When you decide to re-run an interview, it’s free to them and it’s your decision, not a product they were upsold at the worst moment of their week.
No paid retakes.
Ever, in any form, for any reason.
Same conditions for everyone.
Every candidate for a role answers the same question set under the same rules.
Results returned free.
Their own levels and gaps, the same day, whatever you decide.
The interview score doesn’t stop mattering on the hire date.
When you hire someone, their interview scorecard goes onto their employee record. At 90 days, and again at six months, their actual performance is assessed on the same five-level scale.
Which means you eventually get the answer nobody in this market currently has: did the people the system scored highly actually do better? Not in general — in your company, in your roles.
If the correlation is weak, you’ll see that. We’d rather show you a number that undermines our own scoring than have you find out in year three.
One scale, from the interview to the review.
Competency Framework
Defines the skills and levels
AI Interview
You are here
Pipeline & Candidates
Where the scorecard lands
Tanca HRM
Performance on the same scale
Skillify
Development for the levels not yet reached
The first round produces evidence. Your interviewers get a shortlist with two competencies already circled, and the rest of the conversation is theirs.
Measured inside the product.
Chuỗi bán lẻ thực phẩm · 22 cửa hàng
First-round hours removed per month, per open role
Công ty logistics · 3 kho miền Nam
Days from application to human interview
Trung tâm tuyển sinh · Contact Centre Agent
Interview completion rate since returning results to candidates
You’ll see the same numbers for your own account from the first round.
Questions people ask about AI interviews.
Does the AI decide who moves forward?
No. It scores, evidences and recommends. Every one of the five outcomes is triggered by a person, and the product states this on every scorecard.
What if a candidate refuses to be recorded?
They’re routed to a human interview instead. Declining doesn’t remove them from your process.
How long are recordings kept?
180 days by default, configurable, deleted automatically. The candidate is told the number before they start.
Who can watch a recording?
Only roles you’ve granted access, HR and the hiring manager by default. Every view is logged.
Is face matching legal here?
Biometric data is sensitive personal data under Vietnam’s law and requires separate explicit consent, which is what the product asks for.
Do you detect cheating?
No. Four integrity signals come back as indicators, one of them as a probability, for a person to review against the recording. They are never a reason to decline anyone on their own.
Do you score personality or cultural fit?
No, and we won’t. Those are proxies for characteristics that have nothing to do with the job.
Do candidates pay for anything?
Never. Not to take it, not to retake it, not to see their results.
How many interviews are included?
Fifty per month. Interviews you’ve already run stay available; running out only pauses new ones.
Does this replace our interviewers?
It replaces the first screening call. Your interviewers get a shortlist with the gaps already identified, which is a better use of them.
Get your first round back.
Structured, evidenced, and reviewed by the people who were going to spend twenty hours on phone calls.
Start free