The number worth pausing on

In a single month, AI was explicitly cited as the cause of 26% of announced US layoffs. A year earlier, that number was 0%.

The figure comes from the Challenger, Gray & Christmas Job Cuts Report, with UBS Global Research noting that the share has pushed year-to-date AI-attributed cuts to 16% of all announcements. At the same time, 42% of corporate respondents in a UBS survey now expect AI to reduce their hiring pipelines — up from 31% in October 2025.

This isn't the same conversation we were having two years ago.

The entry-level squeeze is the visible edge of a bigger shift

Anthropic CEO Dario Amodei has warned that AI could affect half of entry-level white-collar roles. Microsoft AI chief Mustafa Suleyman put the timeline at 18 months. An Anthropic study found that current systems can theoretically automate most tasks in management, business, finance, and law — and most of those tasks happen to be the ones traditionally given to first-year hires.

This isn't a forecast. The US unemployment rate for 16–24-year-olds has crept up from around 8% pre-2023 to 9.5% this April. The pressure isn't catastrophic yet, but it's directional, and the layoff-attribution data is the leading indicator.

For Italy, the parallel signal is already on the table: entry-level hiring is down 18.8% year-over-year, with the education sector hardest hit at -31.2% (LinkedIn First Job Barometer, April 2026).

What was the entry-level value proposition, exactly?

Junior hires used to win on three things: low cost, rote-task throughput, and trainability. AI now beats them on the first two. Trainability remains — but only if a company can identify it before the interview.

The selection process most companies use today wasn't built for that. The CV evolved as a credentialing document — a way for a recruiter to skim someone in six seconds and decide whether to spend thirty minutes more. It works when the labor market has more roles than candidates. It loses traction when AI inverts that ratio.

The result is a generation of graduates competing on a document format that says almost nothing about what they can actually do.

Where the signal moves next

When the document layer compresses, employers will keep hiring — they'll just shift toward signals that are harder to fake:

Verified work, not claimed work. Projects, theses, code, design files — read by software, not skimmed by humans.
Light institutional vouching. A career office that has seen a student work over three years carries signal that no recommendation letter captures.
Soft-skill evidence, not soft-skill bullet points. What a student is actually energized by versus what drains them, recorded over time, beats "team player" in a list.

A wave of new tools — Poppin' Jobs in the US, several in Europe — is trying to rebuild the entry-level layer. Most are improving the résumé itself, which we think misses the point. The résumé isn't the bottleneck; the absence of structured evidence underneath it is.

What graduates can do now

If you're entering the market this year, three moves compound:

Build a portfolio that contains the work, not summaries of the work. A linked GitHub repo, a thesis PDF, a design file with comments — these are scannable by AI in a way a CV bullet isn't.
Make yourself searchable on what you actually did. Skills extracted from real projects rank differently from skills typed into a profile. The first you can defend; the second you can't.
Show what drives you, not just what you know. As entry-level work gets more automated, the differentiator shifts from "can do" to "wants to do." Recruiters who can see motivation patterns will pay attention to candidates who have made them visible.

What we're building

InTransparency exists for this transition. Students upload real projects; our AI extracts the skills they actually demonstrated; institutions optionally endorse; employers read structured evidence instead of self-declared bullet points.

We don't think the CV will disappear in the next 18 months. We do think it's losing its grip on the entry-level signal — and that the graduates who position themselves on evidence now will compound an advantage on the ones who don't.

The data isn't gentle, but it isn't fatal either. The opportunity is to use what's happening to the labor market as a reason to build a portfolio that lasts — one the next decade of AI doesn't make obsolete.