Entry-Level AI Jobs Are 3% of the Market: 93 of 2,852 Open Roles (September 2026)

On 7 September 2026 the board held 2,852 live AI roles. Exactly 93 of them were explicitly entry-level or internships: 55 internships and 38 entry-level or new-grad roles. That is 3.3% of the market. Mid-level and senior roles together were just under 69%. This is the first monthly edition of the entry-level share, frozen at publication, with the method at the bottom and the live count always on the entry-level AI jobs page.

The seniority ladder, top to bottom

We infer seniority from the title and the text of each posting. The full board on the day of the read:

• Mid-level - 1,369 (48%)
• Senior - 586 (21%)
• Staff - 383 (13%)
• Manager - 227 (8%)
• Principal - 141 (5%)
• Intern - 55 (2%)
• Entry-level - 38 (1%)
• Director - 37
• VP - 15
• C-level - 1

The staff and principal tier alone, 524 roles, is more than five times the size of the entry-level and internship tier. There are more open director, VP and C-level AI roles (53) than there are entry-level roles (38). AI hiring in September 2026 is a market that buys experience it does not want to grow.

Who is actually hiring juniors

The entry-level page renders its 50 newest listings, so the rest of this section is about that sample of 50, not all 93. Palantir held 10 of the 50, every one a Forward Deployed Software Engineer new-grad or internship seat, spread across Chicago, Washington DC, Honolulu, London, Seoul and Sydney. Epic Games held 3, all machine learning internships (London, Montreal and a multi-location listing). Hewlett Packard Enterprise, Skydio, Composio, Mistral AI and Block held 2 each. The remaining 27 listings came from 27 different companies with one seat apiece.

By function, the 50 split into Applied AI 20, ML Engineering 13, Research Science 10, Data & MLOps 3, AI Product 3 and AI Governance & Policy 1. Twenty-nine of the 50 were internships and 11 carried "new grad" or "early career" in the title. The Forward Deployed Engineer title, which barely existed two years ago, accounts for 16 of the 50 on its own once Palantir, Retell AI, LiteLLM, Composio, Endeavor and Mistral AI are added up. If you are a new graduate who wants into AI, the door with the most traffic is deploying somebody else's model for a customer, not training one.

Where the seats are

The sample is US-heavy, with San Francisco and Chicago the most common single cities, but it is more international than the board as a whole. Mexico had three listings (Scale AI in Mexico City, Hewlett Packard Enterprise in Mexico City, Intel in Guadalajara), Paris two (both Mistral AI internships), Seoul two (Palantir), London two, Canada two, and one each in Warsaw, Galway, Zurich, Berlin, Barcelona, Stuttgart, Hyderabad, Singapore and Sydney. A pattern worth knowing: several enterprises place their junior AI seats in lower-cost hubs (Guadalajara, Warsaw, Hyderabad, Mexico City) while their senior AI seats stay in the US.

Remote is close to zero. One of the 50 was fully remote, Abnormal AI's Machine Learning Engineer I in Singapore. Compare that with the board overall, where 378 roles are fully remote. Employers that will hire a junior want that person in the room.

What the ones that publish pay actually pay

Nine of the 50 published a band. The new-grad Forward Deployed Engineer roles are the clearest signal: Retell AI at $150K to $260K, LiteLLM at $150K to $210K and Endeavor's early-career seat at $120K to $140K. Internships that disclosed pay ran from OpusClip's AI Product Management Intern at $25 to $28 per hour, through Epic Games' Machine Learning Intern at an annualised $111,223 to $133,468, to Modal's ML Research Intern at $15K per month. Pallet's Associate Deployment Strategist opened at $100,000. The other 41 listings said nothing about pay, which is a lower disclosure rate than the board's overall 49%.

Two listings that should not be here, and why we leave them

Honesty about the classifier: two of the 50, Abridge's Associate General Counsel, Product and LMArena's Associate General Counsel, AI Product & Privacy, are senior legal roles that landed on the page because the title contains "Associate". We publish the count as classified rather than hand-editing it, so that next month's number is comparable with this one, and we say so here instead. Read the 93 as about 90 real junior seats.

What to do with this if you are starting out

• Internships are 59% of the junior market. If you are still in a degree, the 2027-start listings were already open in September 2026: Skydio for summer 2027, Kodiak Robotics for winter 2027, Phonic for 2027. The window for a 2027 start is now.
• Forward deployed is the volume door. The role is customer-facing engineering: integrating a model into a customer's systems and making it work. It rewards shipping and communication over research depth, which is why it is the one AI title that hires new graduates in numbers.
• Show the artifact, not the coursework. A reproduced paper with the gaps written up, a small eval harness with results, or a documented retrieval or agent project with its failure modes is what these screens read. There is no certification economy in AI hiring to substitute for that.
• Plan for an office. One remote seat in 50.

Method and caveats

Counts are live roles on 7 September 2026, read hourly from each company's own applicant tracking system and classified by title and description. Seniority is inferred from the title and text, so a posting that says "entry level" in the body but carries a senior title is filed by its title, and vice versa. The 93 is the whole board; the company, city, pay and remote breakdowns above use the 50 newest listings the page renders, and say so. It is a large sample of companies that publish a feed, not a census, so treat every count as a floor. Every figure is free to quote with attribution to AI Hiring Board.