What gets lost when AI handles your first-round read.

AI screening tools are good at the structured part of a candidate. They can extract experience, parse qualifications, score keyword overlap with a job description, predict likelihood of matching to past hires. The structured part of a candidate is what fits on a CV.

What gets lost is everything that does not fit on a CV. And the lost stuff happens to be the part that determines who actually makes a good hire.

What a recruiter sees that a model does not.

When a recruiter reads a CV, they are doing pattern-matching, but they are also doing something else. They are noticing the shape of the candidate's narrative. The career story has a coherence or it does not. The progression makes sense or it does not. The choice to take a step sideways, or down, or to leave a company, has a reason or it does not, and the reason is sometimes visible in the way the CV is written, in the gaps, in the choice of how to describe a role.

A model can score the CV's content. It cannot read the narrative, because narrative is not a feature you can extract. The model assigns a probability that this candidate's profile matches the role, and that probability is calculated by comparison to past candidates. It is incapable of evaluating coherence.

What a video moment carries that a CV does not.

When a candidate answers a 60-second video question, the recording carries information the CV cannot. Tone of voice, pace, word choice, eye contact, the moment of hesitation before answering. None of those are signals that get optimised away by polishing the CV. None of them survive a translation into structured data, because there is no clean structure to translate them into.

A recruiter watching a 60-second video moment processes all of this in parallel and without explicit reasoning. The recruiter does not consciously check tone, pace, and word choice. They form an impression. The impression is informed by years of pattern-matching across previous candidates, and it is usually a useful signal.

It is also useful in the cases where the CV does not match the role and the candidate's video moment changes the recruiter's mind. The recruiter reads the CV, thinks no, watches the video, thinks maybe yes. The reverse also happens. Both directions are signal, and both are inaccessible to a model that only sees the CV.

Why this matters for the early funnel.

AI screening tools are typically used at the top of the funnel, where the volume is highest and the bottleneck is most acute. The argument for the tool is that the recruiter does not have time to read every CV, so a model triages the pile and the recruiter only reads the top of the stack.

The cost of that argument is that everything that does not show up in the CV gets filtered out before the recruiter sees it. The candidate who is unconventional, the candidate whose CV is weak but whose presence on a call is strong, the candidate whose career story does not match the model's training data, all of them are filtered out at the same step. The funnel narrows on the dimension the model can measure, which is also the dimension that produces the most predictable hires and the least interesting ones.

Predictable hires are not bad. Most hires should be predictable. The issue is when the predictability becomes the whole funnel, and the unexpected hire never gets a chance to surface.

The shape of the alternative.

The argument is not against AI doing useful work in recruitment. It is against AI deciding who gets read. The shape of a healthier first-round is one where every applicant gets a thin layer of additional context (a CV plus a 60-second video moment, say) and the recruiter reads through that layer faster than they would read through CVs alone.

A 60-second video moment is faster to process than a 600-word cover letter. A recruiter can watch ten of them in ten minutes and form a clear sense of who they want to take to a screening call. The video moment is not a substitute for the CV, it is a complement, and the combined signal is richer than either on its own.

What the recruiter still has to do.

A video moment does not make the recruiter's job easier in the way an AI screener claims to. The recruiter still has to watch, still has to think, still has to decide. What the moment does is give them better material to work with.

Recruiters who have used the format for a while describe it as the difference between reading a manuscript and meeting the author for five minutes. The reading is still useful, but the meeting is what makes you sure.

Flipbase video moments are not a replacement for CV review. They are the slice of context that makes CV review actually useful. The recruiter still reads, still watches, still decides, with material that is fuller than what the CV alone provides.