When Hiring Gets Faster, Judgement Becomes the Entire Interview

Two numbers from two different labour markets point at the same shrinking space: the room where a career used to begin by doing ordinary work, badly, in front of someone who could correct it.

The CV stopped being proof of anything

Morgan McKinley's 2026 Global AI Report, published in October, found that half of Singapore employers now use AI to screen CVs and review interview video. It also found that close to two-thirds of Singapore candidates use AI to write or refine their cover letters, and nearly half say they already optimise their CVs for the systems doing the screening.

Read one way, this is efficiency: hiring gets faster on both sides. Read another way, it is the quiet end of the CV as a useful signal. When the document and the filter reading it are both AI-assisted, the document stops telling anyone what the candidate can actually do. It tells you how well they, or their AI, can describe it.

Hong Kong shows where that leaves the youngest candidates

Hong Kong's Labour and Welfare Bureau reported in September 2026 that entry-level job vacancies suitable for recent graduates have fallen by roughly 60 percent since 2022, even as the city's overall labour shortage forecast narrows. The jobs that used to let someone prove they could do ordinary work are the ones disappearing fastest. Put the two findings together and the pattern holds across both markets: the screening layer is becoming more AI-mediated at the exact moment the entry-level roles that used to generate a track record are vanishing. A candidate increasingly cannot out-write the filter, and, increasingly, cannot hold a junior role long enough to prove the filter wrong.

What still cannot be optimised for a screen

Across the hiring and assessment systems we build, for employers on one side and training programmes on the other, the same thing keeps surfacing: what does not show up in a CV, AI-polished or not, is whether someone, given an ambiguous problem, will go find out what is actually wrong before proposing a fix. Whether they notice friction nobody assigned them to notice. Whether they can take a half-finished system and leave it easier for the next person to work on, not harder. None of that is credential-shaped. It does not optimise for an applicant tracking system, because it was never meant to be read by one. It shows up only in something built, something shipped, something that can be pointed at and argued over.

This is not an argument that credentials or CVs stop mattering. Employers still need a first filter, and AI-assisted screening is not going away because it is imperfect. It is an argument that the filter tells you less at the same moment the alternative, demonstrated work, is the one thing an AI-optimised document cannot fake.

The shift this forces, quietly

The people whose careers compound from here are not necessarily the most credentialed. They are the ones who can point to something they built, debugged, or improved, and explain the judgement behind the decisions inside it. That is a different kind of asset than a well-worded CV, and it’s an asset most hiring pipelines and training programmes aren’t yet built to surface.

Training providers face a version of the same problem from the other side. A programme that ends in a certificate answers the old question: did this person complete the material. It does not answer the question hiring managers increasingly need answered, because the CV cannot answer it either: can this person be handed an ambiguous problem and produce something a team can build on. Programmes built around project work that mirrors a real job, reviewed by someone who can tell the difference between a correct answer and a well-reasoned one, produce a different kind of evidence than a transcript. It is the same gap explored in a related piece on what happens to the people responsible for building a graduate’s judgement once the tasks that used to build it disappear: not the task being automated, but the absence of anyone assigned to replace what it taught. That evidence travels further precisely because it was never meant to be read by a screening algorithm in the first place.

The organisations and institutions that adjust fastest will not be the ones with the most polished CV-writing workshops. They will be the ones that find a way to let people demonstrate judgement before anyone has agreed to hire them, and that count a finished piece of work as more legible than a finished course.

An open question worth sitting with

If the screening layer keeps getting better at filtering documents and worse at measuring the thing documents were always a proxy for, the organisations that keep winning talent will be the ones that built an alternative way to see the work itself.

Is that something most companies and training providers are set up to do, or is it still easier to hire for the CV and hope the judgement was in there somewhere?

‍

‍