Only 1.5% of Agentic AI Jobs Are Junior or Entry Level

We track every agentic AI listing we can find and score it by seniority. As of today the corpus holds 1,719 live roles across 486 companies. Fifteen of them are tagged junior. Ten are internships.

That is 1.5% of the entire market.

What the distribution actually looks like

The full seniority split across 1,719 listings:

• Mid level: 754 roles, 43.9%
• Senior: 401 roles, 23.3%
• Staff: 322 roles, 18.7%
• Lead: 165 roles, 9.6%
• Executive: 52 roles, 3.0%
• Junior: 15 roles, 0.9%
• Intern: 10 roles, 0.6%

Read that from the bottom and the shape of the problem is obvious. There are more executive roles than junior and internship roles combined, by a factor of two.

It is worth pausing on how unusual that is. In a normal engineering job market the seniority pyramid has a base: more junior roles than staff roles, because companies expect to grow people. This pyramid is upside down. The market is buying finished engineers and not making any.

This is a pipeline problem, not a hostility problem

It is tempting to read 1.5% as employers refusing to train anyone. The likelier explanation is duller. Agentic AI teams are mostly small, mostly new, and mostly formed by pulling experienced engineers off other teams. A team of six that formed nine months ago has no mentorship capacity, no onboarding path, and no one whose job is to grow a junior. So it hires mid level, because mid level is the cheapest seniority that arrives self-sufficient.

You can see the same logic in what mid level actually means here. At 43.9% of listings it is not the middle of the distribution, it is the floor. Companies have set their entry bar at the first level that does not require supervision, and then hired most of their headcount there.

That is a temporary condition rather than a permanent one. Every technology wave has looked like this early: no established pipeline, no agreed curriculum, and hiring managers who cannot yet describe what a good junior looks like. It resolves when the work becomes routine enough to delegate. It is also cold comfort if you are trying to get in this year.

Where the junior roles actually are

Here is the finding that changes the search strategy. We cross-referenced the 15 junior listings against employer size, measured by how many roles each company currently has open.

Ten of the fifteen junior roles are at companies with exactly one open position.

Two thirds of the entry-level market sits with employers who are hiring a single person. Meanwhile the ten largest employers in our corpus, who between them account for 44% of all listings, are hiring almost no juniors at all.

This inverts the obvious approach. The companies with the most openings are the least likely to have one for you. The companies nobody has written about are where the entry-level market actually lives, and there are 376 of them.

Three routes that still work

Enter through an adjacent role at a company that already runs agents. The 43.9% of listings at mid level are not all agent-specialist roles. Plenty are backend, data or platform positions on teams that happen to own agent systems. Getting hired next to the work is materially easier than getting hired into it, and internal moves do not compete against the same applicant pool. Six months of proximity also fixes the experience problem that kept you out.

Target the companies with one open role. 376 of the 486 companies in our corpus have exactly one listing, and as above, that is where two thirds of junior roles are. Small teams hire on evidence rather than on process, and a single-opening company is far more likely to read your project than to filter you on years of experience. Those employers also disclose salary less often, at 23% against 44% market-wide, so expect to raise compensation yourself rather than reading it off the advert.

Make the seniority question irrelevant. The reason mid level dominates is that it signals self-sufficiency. Anything you can show that proves you ship without supervision does the same job as a line on a CV. A deployed agent with an evaluation suite and a written account of what broke is worth more than the year of experience you do not have yet.

The last one deserves emphasis, because it is the only route entirely within your control. Hiring managers screening for mid level are not really screening for years. They are screening for the risk that you will need someone to hold your hand for six months. Every artefact that reduces that perceived risk moves you up a level in the reader's mind, regardless of what your CV says.

What we would build

One agent, deployed publicly, that does something narrow and real. An evaluation suite of twenty or so cases that scores its behaviour rather than asserting exact output. Tracing, so you can answer which step failed. And a written account of the three things that broke when it met real inputs, and what you changed.

That last document is the one that does the work. Anybody can follow a tutorial to a working demo. Very few candidates can describe a failure precisely, which is exactly the skill the job consists of.

What to watch

We will re-run this measurement quarterly. The number to watch is not junior share on its own but junior share against mid level. If mid level starts falling while junior rises, teams have begun building pipelines. If both hold steady, the market is still absorbing experienced people faster than it grows new ones.

The second thing to watch is internships specifically. Ten internship listings across 486 companies is close to zero, and internships are the leading indicator for entry-level hiring generally. They are cheap, they are low-risk, and they are how companies discover they can train people. When that number moves, junior hiring follows a year later.

Browse the current openings by seniority on the full job index, or by specialisation on the role hubs. Our take on the single-opening employers is in the long-tail piece.