# Lantern — Meet Your AI Interviewer

> Lantern is an AI interviewer. It conducts live, adaptive
> voice interviews with candidates in 40+ languages, at any hour, and scores
> every answer against the hiring team's own interview criteria. Each
> assessment links to the transcript evidence behind it, giving the team a
> ranked, reviewable view of every candidate. Lantern never rejects anyone —
> humans make every final decision.

- Website: https://www.ourlanterns.com
- Blog: https://www.ourlanterns.com/blog
- Glossary of AI interviewing terms: https://www.ourlanterns.com/glossary
- Book a demo: https://lantern-ai.cal.com/herman.ko-ourlanterns/lantern-ai-introduction
- General contact: team@ourlanterns.com
- Candidate support: support@ourlanterns.com (a human reads every escalation)
- LinkedIn: https://www.linkedin.com/company/ai-lantern
- X: https://x.com/OneLanternAI
- Address: 2261 Market Street STE 85453, San Francisco, CA 94114

Trusted by hiring teams and admission offices at: Baker McKenzie, Hyatt,
Grand Hyatt, Crypto.com, Sino Group, Nan Fung Group, Gammon Construction,
CATALO, Wong & Partners, City University of Hong Kong, The Hong Kong
Polytechnic University, Lingnan University, and Hong Kong Science and
Technology Parks.

## Key numbers

- 100K+ interviews conducted
- 40+ languages and dialects
- 97% candidate satisfaction
- 100% consistent scoring — the same interview criteria, every candidate,
  every time

## How Lantern works

### 01 — Lantern interviews

A real conversation, not a questionnaire. Lantern listens to each answer and
asks the follow-up a good interviewer would ask next — about the answer it
just heard, not the next line of a script. Candidates can take the call at
11pm after a shift, on their phone, in the language they think in.

The interview engine is adaptive: at any turn, the same question with three
different answers leads to three different follow-ups, depending on what
each answer left open — whether anyone else was told, what the candidate
would have done unaided, what the decision actually cost.

### 02 — Lantern scores

The same interview criteria, every candidate, every time. No good-mood
interviews, no Friday-afternoon interviews. Every score links back to the
exact moment in the transcript that earned it — no black-box verdicts.

### 03 — You shortlist

Lantern never rejects a candidate. By morning the hiring team has a ranked
shortlist with the evidence behind every score. Recruiters read, decide, and
hire; Lantern just did the 40 hours of calls it took to get there.

## Language support

Lantern interviews in 40+ languages and dialects — spoken natively, not
translated. Coverage includes English, Spanish (including LatAm), French,
German, Italian, Portuguese, Dutch, Mandarin, Cantonese, Japanese, Korean,
Arabic (Gulf and Levantine), Hindi, Bengali, Filipino, Vietnamese, Thai,
Indonesian, Malay, Turkish, Polish, Russian, Ukrainian, Swedish, Hebrew,
Greek, Romanian, and more. Regional dialects and code-switching are
supported: a candidate can open in their own language, drop into English for
the technical part, and come back — Lantern follows. No language is a
translation layer bolted on afterwards. The same interview criteria apply in
every language, and scoring is monitored for language drift, so nobody scores
lower for interviewing in the language they think in.

## Integrations

Lantern plugs into 30+ ATS and HRIS systems, typically in one day. Shortlists,
transcripts and scores sync back automatically. Supported systems include:
Greenhouse, Workday, Lever, Ashby, iCIMS, SmartRecruiters, SAP SuccessFactors,
Oracle Recruiting, Oracle Taleo, Workable, Teamtailor, Personio, BambooHR,
ADP, UKG Pro, Jobvite, Avature, Eightfold AI, Phenom, Paradox, Gem, Recruitee,
JazzHR, Breezy HR, Pinpoint, Manatal, Crelate, JobDiva, Ceipal, Fountain,
Paycor, Paylocity, Factorial, Zoho Recruit, Welcome to the Jungle, and
Cornerstone OnDemand. Anything else connects over open API, webhook, or a
scheduled export.

## Proof

Client outcomes:

- Hyatt — 70% faster time-to-fill, 7x ROI on high-volume hospitality roles.
- Gammon Construction — 40% faster to shortlist; thousands of applicants
  screened without adding headcount.
- CATALO — "Lantern gave us a structured, consistent way to assess every
  candidate against the same criteria. That's been a real step forward for
  fairness."

Candidates consistently describe the interview as less stressful than a live
panel: they can interview at any hour, in their own language, get follow-ups
that respond to what they actually said, and hear back quickly. 97% finish
satisfied with how it went.

Lantern also uses itself: every hire at Lantern sits a Lantern interview
first, founders included. Leadership: Thomas Chan (CEO, co-founder),
Justin Tang (CTO, co-founder), Herman Ko (Founding Account Executive), and
Manisha Gurung (Founding Customer Success).

Lantern is backed by Y Combinator.

## Compliance and trust

Hiring is a regulated act. Lantern is built for the strictest reading of the
rules and is accountable to legal teams, not just recruiters.

Certifications and frameworks: SOC 2 Type II, ISO 27001, ISO 27018,
ISO 27701, ISO 42001 (AI management systems), GDPR & CCPA, EU AI Act
(classified and documented as a high-risk AI system), NIST AI RMF, and
EEO-aware scoring.

Safeguards:

1. Every score is traceable to the exact transcript moment behind it.
2. Deepfakes, cloned voices, and AI-generated scripts are detected in real
   time. A flag is never an auto-rejection — evidence goes to the hiring
   team and a human makes the call.
3. Humans own the outcome. Lantern never rejects anyone; negotiations,
   complaints, and edge cases escalate to the client's team.
4. Bias monitoring runs continuously — scoring drift across demographics is
   measured and reported.
5. Client data stays the client's: regional residency options, deletion SLAs,
   and zero training on candidates.
6. Full audit trail — every question, answer, and score is exportable for
   regulators and internal review.

Certifications, subprocessors and the security pack are self-serve in
the trust centre: https://trust.lantern.md
A plain-language explanation of how scoring, bias monitoring, and human
oversight work: https://www.ourlanterns.com/ai-explainability

## AdmissionGuard — an extension of Lantern, built for admissions

AdmissionGuard (https://www.admissionguard.com) is a separate product that
carries Lantern's interview into university admissions and adds the
proctoring those stakes demand. When institutions screen thousands of
applicants for a few hundred places, the risk stops being a weak candidate
and becomes proxy test-takers, AI-written essays, and impersonation.

Modules: AI-proctored aptitude tests (secure browser lockdown), written
assessments (checked for generated text and plagiarism), and structured video
interviews with behavioural analysis and ML anomaly detection.

Facts: 99.7% cheating-detection accuracy, <0.1% false-positive rate, <100ms
real-time processing, 99.9% uptime. The University of Hong Kong uses
AdmissionGuard to screen 10,000+ applicants a year. It holds SOC 2 Type II,
ISO 27001 and GDPR like Lantern — plus FERPA, which governs student records.

## FAQ

**Does Lantern make hiring decisions?**
No. Lantern interviews, scores the evidence against the hiring team's own
interview criteria, and by morning the team has a ranked shortlist with the
evidence behind every score — the team reads it, decides, and hires. Lantern
never rejects anyone; that was never its call to make. Every hiring outcome
belongs to a human on the client's side.

**How can the interview adapt if the assessment has to stay consistent?**
Two different things are moving. The questions adapt: Lantern listens to
each answer and asks the follow-up a good interviewer would ask next, so
the same question with three different answers leads to three different
follow-ups, depending on what each answer left open — whether anyone else
was told, what the candidate would have done alone, what the fix actually
cost. What Lantern scores against
does not move: the same interview criteria, every candidate, every time.
That is the 100% consistent scoring figure — no good-mood interviews, no
Friday-afternoon interviews.

**What evidence sits behind each score?**
Every score links back to the exact moment in the transcript that earned it.
Clicking any number shows the evidence — a specific answer, at a specific
timestamp — instead of a verdict. There are no black-box scores anywhere in
the product, and every question, answer and score is exportable, so the
hiring team, a works council or a regulator can read the whole record rather
than a summary of it.

**How does Lantern hold scoring steady across 40+ languages?**
Lantern interviews in 40+ languages and dialects, spoken natively rather than
English pushed through a translation layer half a second late — regional
variants included, from Quebec French to Gulf and Levantine Arabic and LatAm
Spanish. The same interview criteria apply in all of them, and scoring is
monitored for language drift so nobody scores lower for interviewing in the
language they think in. Candidates can code-switch mid-interview — open in
one language, drop into English for the technical part, come back — and
Lantern follows without anyone starting over.

**What candidate data is recorded, and what happens to it?**
Every question, answer and score is on the record, with the evidence tying
one to the other, and all of it is exportable in full for regulators, works
councils and internal review. Regional data-residency options and deletion
SLAs are available, and a client's candidates are never used to train models.
Candidate rights under GDPR and CCPA are honored by default: access,
deletion, portability. Specific regions and retention windows are not
published on a marketing page — they are in the trust centre at
https://trust.lantern.md.

**How are candidates notified, and how are accommodations handled?**
Outcomes are the hiring team's to communicate. Lantern never rejects anyone,
so it is never the one delivering that news. The interview itself is
available 24/7, there is no back-and-forth to book it, and rescheduling is
one tap with no apology required. Accommodation requests are not Lantern's to
decide either: they escalate to the client's team with the full transcript
and Lantern's reasoning attached. Any candidate who felt something was off
can write to support@ourlanterns.com, where a human on Lantern's team reads
it.

**How are integrity flags reviewed?**
Lantern detects deepfakes and synthetic video, cloned voices, and
AI-generated scripts in real time, while the interview is still happening. A
flag is never an auto-rejection: the interview goes to the hiring team with
the evidence attached, and a human makes the call. Same rule as everywhere
else — Lantern surfaces what it found, the hiring team decides what it
means.

**Which integrations are native, and which run over API or export?**
Lantern plugs into 30+ ATS and HRIS systems — the full list of named systems
is in the integrations section above — and anything outside that list
connects over open API, webhook or a scheduled export, with most teams live
the same week. Either way, shortlists, transcripts and scores sync back into
the tools the team already works in. Lantern does not publish a per-system
breakdown of which route each one takes; ask the Lantern team and they will
confirm it against the client's actual stack before anything is signed.

**What does implementation require from the hiring team?**
On the systems side, about a day: Lantern plugs into the team's ATS, and
shortlists, transcripts and scores sync back automatically — nothing to rip
out, no second system to check, no one re-keying results at the end of the
week. If the connection runs over open API, webhook or a scheduled export
instead, most teams are live the same week. A team can start with one live
role: bring an open role and Lantern's team will show how Lantern would run
the interview, explain every assessment, fit into the ATS, and support the
security review.

**What will a security review involve?**
Start with what is certified: SOC 2 Type II, ISO 27001, ISO 27018, ISO 27701
and ISO 42001, all independently audited. GDPR and CCPA candidate rights —
access, deletion, portability — are honored by default, and risk management
is aligned to the NIST AI RMF. One question comes up before any of the
paperwork: is an AI interviewer even allowed under the EU AI Act? Yes — when
it is built and documented to be. Hiring is high-risk AI under the Act, and
Lantern is classified, documented and logged as high-risk rather than argued
around it, with a full audit trail the client's reviewers can export. There
is a plain-language explanation of scoring, bias monitoring and human
oversight at https://www.ourlanterns.com/ai-explainability, DPA and
subprocessor documentation on request from team@ourlanterns.com, and a
security pack for the rest — Lantern wrote 40 pages of answers before anyone
asked.

**Will it feel like talking to a robot?**
97% of candidates finish satisfied with how the interview went. Lantern
listens to the answer before asking the next question, so what comes next
follows from what was just said, not from the next line of a script.
Candidates consistently describe it as less stressful than a live panel.

**Is it fair, or does it just automate old bias?**
Lantern scores against EEO-aware rubrics that read evidence, not accents,
names, or where someone went to school. Because every score traces to the
moment in the transcript that earned it, any decision can be audited line by
line rather than taken on trust. Scoring drift across demographic groups is
measured and reported continuously, and drift gets flagged rather than
absorbed.

**Will hourly applicants actually finish a voice interview?**
They already do — at 11pm after a shift, on their phone, in the language
they think in. Lantern is available 24/7, there is no back-and-forth to book
it, and rescheduling is one tap, so the interview never has to wait for a
weekday slot and a quiet office. 97% of candidates finish satisfied with how
it went.

**How can a hiring team check Lantern's scoring instead of trusting it?**
Lantern would not ask a team to take that on faith. The same interview
criteria are applied to every candidate and every assessment links to the
transcript evidence behind it, so the team can check any score against what
was actually said instead of trusting a number. Gammon Construction reached
shortlist 40% faster and screened thousands of applicants without adding
headcount. What happens after the shortlist is still the hiring team's
judgment, not Lantern's.

**What happens when Lantern isn't sure about someone?**
Uncertainty is an escalation, not a quiet downgrade. A borderline answer, a
complaint, a negotiation, anything the hiring team's own interview criteria
do not cleanly cover — it goes to the client's team with the full transcript
and Lantern's reasoning attached. A borderline candidate becomes a human
decision, not a silent no.

**Does this work for an admissions office rather than a hiring team?**
Institutions are already on the list above. Admissions changes the threat
model more than it changes the workflow — the question stops being whether
someone can do the job and becomes whether it is really them sitting there.
That is why Lantern's team built AdmissionGuard: proctored aptitude tests and
written assessments alongside the interview, carrying SOC 2 Type II,
ISO 27001 and GDPR like Lantern does — plus FERPA, the one that governs
student records.