The highest-paying AI jobs in September 2026
Most lists of the highest-paying AI jobs come from year-old salary surveys. This one comes from 3,758,123 job postings collected in the first half of September 2026, 1,034,726 of which advertised a salary. Ranked by what employers actually offered, chief executives lead at $215,000 and lawyers come second at $200,000 - software developers are fifth. The occupations paying most for AI are largely not the ones using it: chief executives show a 72% premium on postings where fewer than one in ten mention AI, while developers, who mention it 39.5% of the time, get 28%. Plus the skills table, where PyTorch pays $179,938 and the generic phrase "artificial intelligence" pays the least.
Dvir Atias
Founder, JobsPipe
Most lists of the highest-paying AI jobs are built from salary surveys that are a year old and a job title away from the truth. This one is built from 3,758,123 job postings collected between 1 and 18 September 2026, of which 1,034,726 advertised a salary. Every figure below is what employers actually offered this month.
It also arrives at an answer the survey-based lists do not: the occupations paying the most for AI skills are largely not the ones using AI. That turns out to be the most useful thing in the data.
The highest-paying AI jobs right now
Rank occupations by the median advertised salary of postings naming at least one AI skill, and the top of the table is not where the listicles put it. Managing directors and chief executives lead. Lawyers come second. Software developers, the job almost every competing list opens with, come fifth.
| Occupation | Median | Top 10% | vs non-AI |
|---|---|---|---|
| Managing directors and chief executives | $215,000 | $325,000 | +72% |
| Lawyers | $200,000 | $300,000 | +28% |
| Research and development managers | $192,787 | $276,650 | +25% |
| ICT services managers | $188,775 | $282,950 | +28% |
| Software developers | $169,538 | $256,000 | +28% |
| Sales and marketing managers | $161,000 | $255,075 | +50% |
| ICT sales professionals | $157,290 | $273,500 | +85% |
| Finance managers | $155,000 | $235,360 | +21% |
| Computer network professionals | $152,900 | $241,800 | +22% |
| Mathematicians, actuaries and statisticians | $149,135 | $240,991 | +30% |
Occupations need at least 300 salaried AI postings in the window to appear. “Median” is the midpoint of the advertised range, converted to USD.
What jobs use AI the most?
That is a different question with a different answer. Mathematicians, actuaries and statisticians name an AI skill in 56.3% of their postings, more than any other occupation we track. Software developers follow at 39.5%. Nobody else is close.
Across the whole market only 6.7% of postings mention AI at all, which is worth remembering the next time someone tells you every job now requires it.
Does AI actually pay more?
Yes, and least where it is most common. Compare the median advertised salary of AI-skilled postings against everything else in the same occupation in the same month, and chief executives show the widest gap in the data: $215,000 against $125,000, a 72% premium, on postings where fewer than one in ten mention AI at all. Sales and marketing managers follow at 50%, HR at 49%, advertising and marketing at 48%. Software developers, the group most likely to use these tools daily, show 28%.
| Occupation | Postings | Name AI | Premium |
|---|---|---|---|
| Mathematicians, actuaries, statisticians | 11,714 | 56.3% | +30% |
| Software developers | 113,721 | 39.5% | +28% |
| Graphic and multimedia designers | 14,884 | 11.7% | +55% |
| Sales and marketing managers | 89,541 | 11.2% | +50% |
| Managing directors and chief executives | 18,884 | 8.8% | +72% |
| HR and careers professionals | 27,899 | 6.0% | +49% |
| Nursing professionals | 48,758 | 1.2% | −5% |
The two columns rank almost in reverse. Where AI is normal, saying so is worth little. Where it is still rare, saying so is worth a lot.
Which AI skills are most in demand?
The skill table makes the same point a second way. The most common AI term in job postings is the plain phrase artificial intelligence, in 45,770 salaried postings, and it carries the lowest median of the leading terms. The specific capabilities pay.
| Skill | Postings | Median salary |
|---|---|---|
| Artificial intelligence (generic) | 45,770 | $114,920 |
| Agentic AI | 17,566 | $170,000 |
| Machine learning | 15,429 | $156,500 |
| Large language models | 13,817 | $165,000 |
| Generative AI | 10,898 | $147,500 |
| AI safety | 5,710 | $164,500 |
| Retrieval-augmented generation | 4,343 | $157,500 |
| PyTorch | 2,820 | $179,938 |
Two things stand out. PyTorch is the best-paid single AI skill in the market at $179,938. And agentic AI has already overtaken machine learning by volume, 17,566 postings to 15,429 — a category that barely existed as a hiring term is now the most-requested specific AI skill.
Why the premium sits outside engineering
The likeliest explanation is scarcity plus mandate. In software, AI fluency is approaching table stakes, so naming it separates a candidate from nobody. In a commercial or executive role it is still unusual, and those postings are generally not describing the same job. A chief executive posting that names machine learning is usually a company hiring someone to lead an AI product line, not a company that wants its CEO writing notebooks. The premium attaches to the mandate, not the tool.
The bottom of the table fits the same reading. Nursing professionals name AI in 1.2% of postings, and those postings advertise 5% less than the rest. Where AI reaches hands-on care work it arrives as scheduling and documentation software, which belongs to more administrative and typically lower-paid roles.
What we are not claiming
These are advertised salaries, not paid ones, and a gap between medians is not a causal estimate. An AI-skilled posting differs from its neighbours in seniority and mandate as well as tooling, and we have not controlled for either.
We did test the most obvious confound. Because salary disclosure skews heavily to the United States, a premium could be a geography effect in disguise. It is not: restricting both sides of every comparison to US postings leaves the gaps essentially unchanged, at 44% for sales and marketing managers and 23% for software developers.
We are also not telling you whether any of this is growing, and we would treat anyone who does with some suspicion. A single month is a photograph, not a direction.
The practical version
Hiring engineers: an AI skill in the job description no longer differentiates your posting, because four in ten competitors have it. Name the specific capability instead.
Hiring commercial or executive roles: it differentiates enormously, and the market is currently pricing that difference at 40–70%.
Writing a CV outside engineering: the scarcest and best-paid position right now is being the person in a non-technical function who can actually run this stuff.
Writing a CV inside engineering: “artificial intelligence” is the lowest-paid thing you can write. PyTorch, agentic systems and LLM work are worth roughly $50,000 more at the median.