Search for the types of entry level AI jobs and you get a list. Seven roles, ten roles, fourteen roles, in no particular order. Annotation sits next to machine learning engineering as if they were the same kind of job. They are not. One is contract task work. The other is a career with a PhD-shaped line in front of it.
So this is the list with the structure put back in. Twelve roles, sorted into six families, and for each one: what you actually do, whether you need code, what it pays where there is a public number, and how open it really is to someone in their first job.
If your real question is how to get hired when you have nothing on your CV yet, we answered that separately in entry level AI jobs with no experience. This page is the map of what exists.
The short answer
There are six families of entry level AI jobs:
- Build: junior AI engineer, junior ML engineer. Code required. Highest pay, fewest true beginner seats.
- Data: data scientist, data analyst on an AI team. SQL and Python. The most common technical door for bachelor's grads.
- Research: research assistant or intern. The biggest category on paper, mostly gated to master's and PhD students.
- Operations and product: AI operations and automation analyst, AI product analyst. Light code. The newest and most open lane.
- Customer-facing: solutions associate, AI sales development rep, AI customer support. Product fluency over code.
- Train, evaluate and govern: AI trainer or annotator, trust and safety or governance analyst. No code. Easiest to start, but the trainer half is mostly contract work.
The pattern that matters: most of these jobs are not titled "AI". And the type the internet points you at first is usually the weakest career start.
How many entry level AI jobs actually exist
Fewer than the listicles imply, and the type you pick decides how long the line is.
Tokn hand-checked every internship and explicit entry-level title on its board and found 67 real entry points in Europe (6.9% of AI jobs) and 95 in North America (3.3%) as of 28 October 2026. OpenAI and Anthropic, two of the biggest employers on that board, listed no internships or new-grad roles at all. Their doors are fellowships and residencies.
Extern ran a bigger sample on US internships. Of 19,533 postings, 4.2% had an AI, machine learning or data science term in the title, rising from 3.1% in August to 6.0% in early November 2026. Of those AI-titled internships, about 36% were research roles, 19% data science, 19% ML or AI engineering, and 14% non-technical: operations, automation, product, strategy and legal.
Read that split slowly. The biggest slice is the one most undergrads cannot enter. The non-technical slice is small but real, and it is where business and humanities grads get in.
Family 1: build
Junior AI engineer
You build product features on top of existing models: retrieval, agents, prompts, evaluations and the plumbing around them. 1752vc calls it one of the faster-growing technical doors for new grads because it rewards building over theory. It needs Python or TypeScript, API work and proof you can measure whether an AI feature works.
Pay: there is no BLS category for AI engineer. The closest benchmark is software developers at a $135,980 median in May 2025. That median covers every experience level, so starting pay sits below it. For a new-grad anchor, the same source cites NACE's projected $81,535 average starting salary for computer science graduates.
The catch: Tokn found AI engineer is rarely a first job title. The real version is a new-grad software engineer role at an AI startup.
Junior ML engineer
Closer to the data and the math: training, fine-tuning, deploying and monitoring models. Python, statistics, an ML framework, some data engineering. Many postings ask for a related degree and some prefer a master's. Nearest BLS benchmark: data scientists, at a $120,230 median.
Family 2: data
Data scientist
You turn data into predictions and decisions. Extern calls it the most common technical AI role open to bachelor's grads, with BLS projecting 35% growth from 2025 to 2035. Statistics, Python, and the ability to explain a result to someone who does not care about your model.
Data analyst on an AI team
You measure how AI features perform with real users: usage, quality, cost, experiments. SQL, a BI tool, basic statistics, clear written summaries. One of the more realistic AI-adjacent first jobs for math, economics and business majors, and a common launchpad into data science.
Family 3: research
Research assistant or intern
Running experiments, reading papers, testing new methods. This is the role people picture when they hear "AI job", and it has the highest wall. Extern notes a master's degree is the typical requirement for research scientists, and many internship postings say PhD in the title. Tokn counted research as 31 of North America's 95 entry points, mostly aimed at PhD or master's students.
If you are an undergrad, treat research as a long shot through a university lab or a fellowship, not as your main plan.
Family 4: operations and product
AI operations and automation analyst
You connect AI tools to real workflows across finance, marketing and operations: automating reports, wiring agents into a CRM, documenting processes, fixing what breaks. Extern calls it the newest and most open lane. Tokn's hand-checked list had 13 AI ops and automation roles in North America and 3 in Europe. Small numbers, but a lane where a business degree is not a disadvantage.
What gets you hired is a before and after: a process that took four hours and now takes one, with the workflow to show it. We go deeper on this role in AI automation analyst.
This is also the family we know from the inside. At EX EPIC Academy, interns are placed in operational roles at live ventures such as Zero X and Gemino AI, with tracks across AI and automation, consulting, engineering, sales and research. The pattern we see every intake: the person who automated one real process for one real team gets taken seriously. The person with three certificates does not.
AI product analyst or associate product manager
You decide what an AI feature should do, who it is for and how to measure whether it works. You do not have to code, but you do have to know what models can and cannot do. New grads usually start as product analysts or associate PMs.
Family 5: customer-facing
Solutions associate
You help customers adopt an AI product: scoping use cases, building demos, answering technical questions next to account executives. It is the junior cousin of the solutions engineer and forward deployed engineer roles. Enough technical depth to build a working demo, plus real comfort talking to customers.
AI sales development rep
Researching prospects, writing outreach and booking meetings for an AI company. Resilience, good writing and genuine fluency with the product. Selling AI well often means explaining what it cannot do.
AI customer support
Helping users with an AI product, diagnosing problems and turning ticket patterns into product feedback. Mercor cites Indeed data putting average US customer support pay at about $19 an hour as of May 2026. Support engineer versions add debugging and API skills and pay more. One honest caveat: 1752vc points out that customer service is one of the case studies in the Stanford research on young workers in AI-exposed jobs, so pick the support roles that sit close to the product team.
Family 6: train, evaluate and govern
AI trainer, annotator or evaluator
You label data, write and grade model responses, and build test cases that check quality. It is the type every listicle leads with, partly because many of those listicles are written by platforms that hire for it.
Pay varies widely. Mercor puts AI trainer work from $15 an hour, exceeding $60 in some specialist cases. DataAnnotation, via 1752vc, lists $25 to $50 an hour for general projects. The structure is the issue, not the rate: it is usually contract work with variable hours. Use it for income and to learn how models fail. Do not mistake it for the first rung of a ladder unless you bring domain expertise (law, medicine, coding, a second language) that moves you to better-paid evaluation work.
Trust and safety or governance analyst
You check AI systems for bias, privacy problems, unsafe output and compliance gaps. Law, policy, philosophy and business majors fit well. Demand is real: a Study.com survey of hiring managers named AI security and risk analyst the most in-demand entry level AI role heading into 2026.
All types side by side
| Role | Family | Code needed | Open to beginners | First proof |
|---|---|---|---|---|
| Junior AI engineer | Build | Yes | Low, mostly as new-grad SWE | A shipped app on a model API |
| Junior ML engineer | Build | Yes | Low, degree often required | A trained, evaluated model |
| Data scientist | Data | Yes | Medium | A modeling project with a recommendation |
| Data analyst, AI team | Data | SQL, some Python | Medium | Dashboards and analysis memos |
| Research assistant | Research | Yes | Very low for undergrads | Lab work, a paper or replication |
| AI ops and automation analyst | Ops and product | Light | High, and growing | A workflow that saves real time |
| AI product analyst | Ops and product | No | Medium | A teardown of an AI feature with metrics |
| Solutions associate | Customer-facing | Some | Medium | Demos built for real use cases |
| AI sales development rep | Customer-facing | No | High | Results from a target-driven job |
| AI customer support | Customer-facing | Light | High | Clear writing, product fluency |
| AI trainer or annotator | Train and evaluate | No | Very high, contract | Careful writing, domain expertise |
| Trust and safety analyst | Govern | No | Medium | A policy memo or model evaluation |
"Open to beginners" is our read of the Tokn and Extern counts above, not a measured score. Pay by role is a whole topic, so we are giving it its own page: how much entry level AI jobs pay.
Which type fits your background
- CS or engineering: build family first, solutions associate as the people-facing alternative.
- Math, statistics, economics: data scientist or data analyst on an AI team, then AI product analyst.
- Business, marketing, operations: AI ops and automation analyst, AI sales development rep, AI product analyst. If you are not sure which skills to show, start with what AI skills actually are.
- Law, policy, humanities: trust and safety or governance analyst, and specialist AI trainer work where your domain earns a higher rate.
- No technical background at all: customer-facing or operations, while you learn AI without coding on real tasks.
Across every type the screen is the same: proof of work over adjectives. That is why we built EX EPIC Academy around shipping, with participants from 26 nations working on 15 live projects. If you want that first piece of proof inside a real company, read our guide to getting an AI internship with no experience.
FAQ
Are there remote entry level AI jobs?
Very few outside contract training and annotation work. Tokn counted only 6 remote roles among its 162 entry points in Europe and North America. We break down where the remote ones are in remote entry level AI jobs.
Is AI engineer an entry level job?
Rarely as a title. The realistic version is a new-grad software engineer role at an AI startup, where you build features on top of models and grow into the AI engineer title.
Which type of entry level AI job is growing fastest?
Agent and LLM work. Extern found 52 US internships with "agent", "LLM" or "generative AI" in the title, a category that barely existed two years ago, and the AI-titled share of internships nearly doubled over the fall 2026 recruiting season. On the business side, Extern calls AI operations and automation the newest and most open lane.
Do big AI labs hire for entry level roles?
Mostly not through normal job posts. OpenAI and Anthropic listed no internships or new-grad roles on Tokn's board in October 2026. Their entry routes are fellowships and residencies, which are small and competitive. Startups post most permanent entry-level AI roles.
