Human-led by design
The difference isn't whether AI is present.
It's whether AI judges the candidate.
Capifiq can help prepare questions, run the interview room, and protect session integrity. It does not conduct an autonomous interview, grade the candidate's answers, rank applicants, recommend an outcome, or make a hiring decision.
A stricter line
Human oversight matters. This goes further.
"A human makes the final call" is a common reassurance, and it is worth less than it sounds. If an automated score reached the decision-maker before they clicked, the score shaped the decision: the click just recorded it.
Capifiq is built to keep candidate evaluation out of the automated layer altogether. The interviewer evaluates the candidate. Capifiq records the conditions the interview happened under. There is no score for a human to defer to, because none is produced.
Where the automation sits
What AI may and may not do here.
AI may assist with
- Drafting interview questions, with a person approving every one
- Source-grounded material for the interviewer
- Captions and transcription
- Media-authenticity analysis of the feed
- Assembling the report
- Scheduling and workflow
AI may not
- Conduct the interview autonomously
- Assess job fitness
- Grade answer quality
- Rank candidates
- Recommend advance or reject
- Trigger an automated adverse action
Two kinds of output
Direct events and analytical signals are not the same claim.
Conflating them is how integrity products end up overstating what they know. Capifiq labels them apart, on the page and in the record.
Direct technical events
Something the platform observed directly: an active remote-control channel, a virtualised environment indicator, a display-topology change, a covered application window, a non-local input event. Recorded with a time, a source, and an evidence reference.
Analytical signals
Model-derived observations about the media itself, whether the feed appears synthetic or internally consistent. These are probabilistic. They are labelled as such, never described as proven, and always routed to a person.
No model-derived signal becomes an automatic accusation or a hiring decision. And the absence of an event is not evidence that nothing happened. It means nothing covered was observed.
Why the line is drawn here
Automating candidate evaluation adds obligations.
Employment, disability, privacy and biometric obligations depend on the jurisdiction and on how a product is used. Systems that evaluate candidates, produce scores, or substantially assist screening can attract additional audit, notice, explainability, accommodation and bias-management duties.
Capifiq is designed to reduce that automated-decision surface by limiting automation to preparation, session integrity and recordkeeping. That is a narrower claim than an exemption, and it is deliberately narrow: customers remain responsible for their own compliance, and nothing here is legal advice.
Questions
Common questions.
Does Capifiq use an AI interviewer?
Does Capifiq score candidate answers?
Does this mean no AI or employment law applies?
What does Capifiq actually report?
Who reviews the record?
Get started
Humans interview. Capifiq verifies.
See the evidence model on your own interviews.