Where entry level went

Nobody is born mid-level. That is the idea CoBlack was built on, and this month the data caught up with it.

A first job should not require a previous one. That sounds obvious enough to be beneath argument. Then you read the postings.

The ladder pulled up its rung

Indeed Hiring Lab economist Jack Kennedy published an analysis on September 17 of millions of US job postings on Indeed that carry an advertised annual salary, comparing the occupations most exposed to AI with those least exposed. The report frames its findings as good news about pay, and they are. “Since 2021, advertised pay in the most AI-exposed occupations has climbed by about 46% (versus 25% in the least-exposed).”

Sitting inside that good news is a different number.

“In the most-exposed occupations, the entry-level share of postings fell from 29% to 10% between 2021 and 2026, while the senior share rose from 22% to 47%.”

Indeed itself notes that part of the pay rise reflects a shift toward fewer, more senior postings in AI-exposed work. Three limits are worth stating plainly. This is postings data from one platform. A posting mix is not a hiring count. And the slide from 29 to 10 happened within the most AI-exposed occupations, not across the whole economy. Even with all of that said, in the corner of the market where the future is arriving first, the first rung is thinning.

Experience was always a proxy

Ask what “experience required” actually asks for. It is not years. It is capability. An employer wants someone who can do the work. Years spent holding a title is simply the cheapest signal to filter on, so the market filters on it.

That shortcut has a cost, and the cost lands in one place. When postings climb the seniority ladder, the proxy screens out people who can do the work but have never held the title. A person is not less capable because no one has hired them yet. The proxy just cannot see them.

A reason that eats itself

ManpowerGroup asked 39,878 public and private employers across 42 countries about their plans for the fourth quarter of 2026, in a survey fielded July 1 to 31 and released September 8. The picture is warmer than the postings suggest. 45 percent of employers say they are adding early-career talent compared with 2025, against 20 percent pulling back. Two limits belong next to those numbers: the survey measures stated intent, not completed hiring, and ManpowerGroup is a staffing company.

Now look at the 20 percent. Among the employers scaling back, the survey found that “cost pressure and a lack of experience weigh more heavily than AI, suggesting automation is reshaping early-career roles rather than eliminating the need for them.”

Read that slowly. A lack of experience is named as a reason not to hire the very people whose first job would be the only way to get experience. The reasoning is circular by construction. The door is closed because you have never been through a door.

Reading the work, not the title

This is the problem CoBlack was built against. We match people to roles on the Career Capability Map, which reads demonstrated capability: projects, coursework, internships, self-taught work, work actually done. Not titles held. Auto Resume writes a resume per opening from that map, and it never inflates past what is true. Auto Search reads validated employer career pages and ATS feeds, so the openings it works from are real.

Here is our honest limit, stated as plainly as the others. CoBlack cannot make an employer open an entry-level role. What it can do is make sure the capability a person does have is visible wherever a door is open.

Entry-level hiring is not a charity line in a workforce plan. It is how every senior person working today got started. Every senior posting is a request for someone who was once given a first yes.