What 8,000 Job Postings Say About AI in Data Roles
We read more than 8,000 live US postings across data roles and industries. Where AI is already table stakes, where it is barely asked for, and what to learn depending on the job you want.
Over the last few weeks we analyzed more than 8,000 live US job postings across data roles and industries, to see how employers are actually asking candidates to use AI. Traditional data skills are still alive and well. AI is becoming an additional layer on top of them, and how thick that layer is depends on both the role you want and the industry you want to work in.
Data scientists cannot learn AI later
AI showed up in almost 9 out of 10 data scientist postings, one of the highest rates we found across roles. If you’re a data scientist or want to become one, stop treating AI like something you can pick up later.
That does not mean skipping the foundations. Machine learning appeared in 83% of data scientist postings, compared with 37% for GenAI and LLMs and 22% for agents. Learn statistics, experimentation, and machine learning properly. Then add GenAI, agents, and RAG. Do not skip the foundations because agents are getting all the attention.
If what you actually want is to be an AI engineer, building agents, RAG systems, evaluations, and production workflows should be much higher on your list.
Only 1 in 4 analyst postings asks for AI. That is a window.
We looked closely at 340 data analyst postings. 91% focused on traditional skills like SQL, BI, and data pipelines. Only about one in four asked candidates to know or use AI at all. Most employers do not yet expect analysts to turn their workflows into automated AI systems.
Where AI does appear in analyst postings, the clearest ask is using it to analyze data, write code, and move faster. So learn to use AI throughout your existing workflow. I would not spend much time on the underlying models unless you want to move toward data science or ML.
If you can do both, be a very good analyst and build reliable systems around your own work, you have a real chance to stand out before everyone is expected to. That is not true for data science and ML, where AI has already become table stakes. Analysts have time to get very good at this before more employers begin expecting it.
Where to start: pick the most boring analysis you repeat on a regular basis and turn it into a reliable agentic workflow. Start with work you know well enough to catch being wrong. Make it show its work, test it against analyses you trust, and fix one failure at a time. The value is not just getting boring things done faster. It’s using the saved time to investigate more questions, catch important changes sooner, and spend more time helping the company decide what to do.
Data engineers and analytics engineers
Data engineers: learn to build reliable pipelines and data products first. Then learn what changes when the consumer of that data is an LLM or an agent. That means unstructured data, retrieval, embeddings, context, and what makes data trustworthy enough for another system to act on it. Spend less time on the models themselves and more on the foundations those models depend on.
Analytics engineers were probably the most interesting. AI-assisted analytics appeared in 36% of postings, more than any other role we analyzed. Focus on semantic layers, clear metric definitions, AI-assisted development, and making company context usable by both people and AI systems. That sample was only 72 postings, though, so treat it as directional.
The industry matters as much as the role
The required skills of a data analyst in tech are not the same as those of a data analyst in health care.
Nearly 60% of data postings in tech and banking asked for AI experience in some form. In health care, insurance, and education it was closer to 30%.
Banking was the only industry where AI appeared more often than expected across data scientists, data engineers, and data analysts. Compared to other roles in banking, the data team had a much higher expectation of AI fluency in incoming hires.
Tech was more evenly distributed across roles. Most roles in tech ask for some AI experience, which we all kind of knew. What surprised me is that 49% of tech data analyst postings asked for AI, roughly double the rate for analysts in other industries. If you’re an analyst somewhere else and want to break into tech, you need to learn how to use AI in your day-to-day workflow.
Health care and insurance sat below average across nearly every role where we had enough postings to compare. I would not read that as those industries being behind. A lot of it is regulatory, particularly in health care.
What to do with this
If you want a data role in tech or banking, expect AI fluency to become part of the job very soon. It already is in many companies. Learn it alongside the traditional skills the role depends on.
If you work in health care, insurance, or education, I would put even more weight on learning how AI fits the workflows you are likely to encounter, because many of your peers aren’t doing it yet. This is a snapshot of what employers ask for right now, not where those industries go next. That gap is a career arbitrage moment.
A lot of people will start with whichever version of AI is getting the most attention. Start with the job to be done instead. Learn the traditional skills that job depends on, then learn the version of AI that helps you do that job better. Sounds simple when you put it that way. I know it isn’t.
The tables, the method and the downloadable data are on the research page for this study.
10+ years in product data science, causal inference and AI evaluation at Stripe, Nextdoor and Ontra.