Once you pause, your history will appear here.
Candidates
Everyone who has completed an assessment for your venue. Click any row for the full breakdown.
Total candidates
0
All-time
Awaiting review
0
Need your decision
Above threshold
0
At current threshold
Avg score
—
All assessed
Threshold
70
Showing candidates scoring ≥ 70
0 below threshold — lower the bar to view
0255075100
Filter, not auto-reject
All candidates
Loading candidates…
Interviews
Candidates you've requested an interview with. Mark hired here once they start.
Loading interviews…
Hires
Active hires. Track who you've brought on board and manage your hiring pipeline.
Loading hires…
Transparency · How TRIAL. scores
How scoring actually works.
No black box. Here's exactly what the AI does, what it sees, what it doesn't see, and where it can be wrong. Read this once and you'll never wonder what a score means again.
The shape of every assessment
3 steps01
Application lands
An applicant applies — Seek, Indeed, Jora, your website, wherever. Their application gets forwarded to your Trial. inbound address and the platform automatically sends them a hospitality-specific assessment link. 10 questions, role-matched, ~15 minutes. No manual step required from you.
02
Candidate answers · we watch how
They answer 10 questions calibrated for their role tier (junior / mid / senior). While they answer, we also quietly track how they answer — paste events, tab switches, time-per-question, answer structure. Two signals running in parallel: substance and integrity.
03
You see a ranked profile
Their answers go to Claude Haiku 4.5 with a role-aware rubric. You get back: a 0–100 score, a 2-paragraph summary, strengths, concerns, risk level — plus the integrity signals separately. You decide who to interview. TRIAL. never makes the hire call.
What the score actually means
0–100Score's calibrated for hospitality, not generic employability. Think of it as a sorting tool, not a deciding tool. Three bands worth memorising:
70 – 100
Worth reading
Specific, considered, hospo-fluent answers. These are the people you want to interview first. Doesn't guarantee they'll be the hire — but they've earned 30 minutes of your time.
40 – 69
Judgement zone
Decent answers with gaps, or strong answers with concerns flagged. Read the AI summary and strengths/concerns. Sometimes a 55 with a great strengths note beats a 72 with a red flag in concerns.
0 – 39
Probably skip
Generic, evasive, or thin answers. Or templated/copy-paste patterns. Doesn't mean the person is bad — means the assessment didn't surface anything worth your interview hours. You can still progress them; the AI just isn't recommending it.
The line we use internally
The platform's job is to save you ~80% of your interview hours, not to replace your hiring judgement. A score gets a candidate in the room. Whether they leave it with an offer is still your call.
The integrity layer · separate from the score
Behavioural signalsSubstance score answers "is the answer good". Integrity answers "did this person actually produce this answer". Two different questions. We surface both because they tell you different things.
Signal
What we capture
What it tells you
Paste events
Browser clipboard API counts paste actions per question
Likely copied from elsewhere — ChatGPT, prepared notes, another tab
Tab switching
Page Visibility API tracks when the assessment tab loses focus
Candidate was somewhere else mid-answer. Not always damning — could be a phone call. Context matters.
Response timing
Time-per-question vs the question's complexity
Too fast = pre-prepared. Too slow with short answers = laboured. Patterns matter more than individual times.
Answer structure
Sentence patterns, vocabulary consistency, voice
Templated vs natural. Same person across 10 questions vs voice shifts.
It is possible for a candidate to score 85 on substance and 40 on integrity — high-quality answers that aren't theirs. We surface that. You decide what to do with it. Worth a follow-up question in the interview; not necessarily a deal-breaker.
AI fairness · what we deliberately don't do
DefensibleWhat the AI sees — and what it doesn't
The scoring prompt receives: the candidate's name, the role applied for, the venue, the written assessment responses, and our role-tier rubric. That's the full input. Deliberately not sent: government ID, payment information, sensitive information (within the Privacy Act meaning), images, audio, video, biometric data, or financial information. Full detail in our AI Transparency Notice §4.
Same rubric applied to every candidate, every time
Questions are selected from a role-specific bank, but every candidate for the same role is scored against the same rubric — the same scoring criteria, the same role-tier lens, the same prompt. No "I had a bad day" bias. No human inconsistency across morning vs evening reviews. The AI's job is consistency. AI outputs are probabilistic — two near-identical answers may score slightly differently (±5 points typical).
Candidates retain their statutory rights
Every assessment email tells the candidate their answers will be analysed by AI. Under the Privacy Act 1988 (Cth), candidates can ask to access the information held about them, correct inaccuracies, or have it deleted — at any time, via privacy@hiretrial.com.au. TRIAL. doesn't undertake to manually re-score on request, but may review in exceptional circumstances at the Operator's discretion (AI Transparency Notice §6). Hiring decisions sit with the venue, not the AI and not us.
Every score is stored and reviewable
Every assessment, score, and AI summary is stored against the candidate record for the duration of the retention window — 6 months from the candidate's last activity. Venues can review any score and its summary at any time from their dashboard. After the retention window, records are deleted in line with the Privacy Policy. Export anything you need before then.
The honest limitations
Things to knowNo assessment is perfect, including this one. Knowing where the AI can mislead you is half the value of using it well.
What the AI catches well
- One-liner answers and "I don't know" patterns
- Copy-paste from ChatGPT or prepared notes (high accuracy)
- Templated, generic, voice-flat responses
- The difference between someone winging it and someone with real shifts behind them
- Specific operational detail vs hand-wavy theory
What the AI can miss
- A great candidate who's bad at writing but excellent in person
- Non-native English speakers whose phrasing reads awkwardly
- Cultural communication style differences
- Genuine introverts who give shorter but accurate answers
- Someone gaming with rehearsed answers (rare but possible)
Consistency, not infallibility
The AI applies the same rubric to every candidate, every time. That consistency is the point — no human reviewer having a bad day, no different standards across a Tuesday morning vs a Friday afternoon. But AI outputs are probabilistic. Two near-identical answers may score slightly differently. Use it as a sorting tool, not a measuring instrument. A 72 is not definitively better than a 70.
The rubric improves over time
The AI model doesn't learn from your data. What improves is the rubric — the instructions we give it. We track retention outcomes and review whether scores are predicting the right things. Where they're not, we update the rubric. Same model, better instructions over time.
A filter, not a guarantee
Nobody can tell you a candidate will work out. Trial. helps you spend your interview time on the candidates most likely to be worth it — and protects you from wasting hours on applications that were never going to work. The hire decision stays with you.
Questions, corrections, or concerns about a score
Your score threshold lives in Settings — drop it to 0 and you'll see every candidate who applied, regardless of what the AI thought. That's the first answer to "is the AI gatekeeping me." Beyond that: candidates have statutory rights under the Privacy Act 1988 (Cth) — access, correction, deletion — at any time via privacy@hiretrial.com.au. Venues with concerns about a specific AI output can email hello@hiretrial.com.au; we don't undertake to re-score on request, but may review in exceptional circumstances at our discretion (AI Transparency Notice §6). Assessment records are kept for 6 months from the candidate's last activity, then deleted — export anything you need for your own records before then.
Settings · Venue
Manage your subscription.
Plan info, automation, and pause / change / cancel — all in one place. Changes save instantly.
Venue details
Venue name
—
Status
—
Forwarding inbox
—
Contact
—
Your plan
Founding PartnerActive subscription
Founding Partner
· Solo
Locked for life · among the first 50 venues on TRIAL.
Monthly
$99.99+ GST
$109.99 inc. GST
Next bill
—
—
Member since
—
Founding cohort
Your Solo pricing is locked for life. As long as your subscription stays active, your monthly rate never increases — even when later venues join at the standard rate.
Candidate visibility threshold
Moved to Candidates
The visibility threshold lives on the Candidates page, right above your candidate list — so you can tune it from where you actually use it. Below-threshold candidates collapse under an expandable row in the table. Never auto-rejected.
Currently
70
Pause subscription
$1 / dayQuiet month? Closing for renovations? Pause your subscription for $1 per day, billed monthly. Your data, candidate pipeline, and locked Founding Partner pricing all stay live. Reactivate anytime.
The math
Pause for
days
Normal monthly rate
$89.99
Pause cost (30 × $1)
$30.00
You save vs full rate
$59.99
Your $1/day pause is
67% off
0 of 90 pause days used this year · 90 remaining
Better than cancelling
Pause keeps your pricing locked. Cancel and re-sign means losing your Founding rate.
Hospo is seasonal. Use pause during Jan, Feb, July. Keeps your data, keeps your price.
Pause history
No pauses yetAutomation
Integrity onlyTRIAL. flags candidates whose assessment behaviour suggests integrity issues — pasted answers, excessive tab switches, or impossibly fast completion. Flags highlight, they don't reject. You make every hiring decision.
High paste-count flag Integrity
Mark candidates who pasted text
3+ times during their assessment. Often indicates AI-generated answers or copy-paste from a prepared script.
High tab-switch flag Integrity
Mark candidates who switched browser tabs
5+ times mid-assessment. Acceptable for some roles, suspicious for others — you decide.
Suspiciously fast completion Integrity
Mark assessments completed in under
40% of the median time for that role. Either a very experienced applicant — or someone speed-running through it.
By design
"TRIAL. never auto-rejects, auto-hides, or makes hiring decisions on your behalf. Every candidate is fully visible to you. These flags are visual cues — you decide what to do with them. This aligns with emerging Australian AI fairness standards."
Cancel subscription
PermanentCancelling ends your subscription at the end of the current billing period. Your candidate data is retained for 90 days, then permanently deleted per our data policy. If you're on Founding pricing, that locked rate cannot be restored if you re-sign later — you'd join at the current standard rate.