AI AND YOUR JOB · CHECKED OCT 2026
Will AI replace data scientists?
No, and demand is still rising. AI notebooks and chat tools now clean data, write Python and draw charts in minutes, but research agents still fail most real analysis tasks, and someone has to pick the question, check the model and defend the result. BLS projects data scientist jobs to grow 35% from 2025 to 2035, partly because companies need people to put AI to work.
- Tasks from the US Labor Department's O*NET
- Pay and outlook from BLS
- 11 sources
AI EXPOSURE TODAY
Medium
AI does 1 and helps with 9 of 13 core tasks. 3 are still human.
- AI does it (1)
- AI helps (9)
- Still human (3)
Task by task: what AI can do today
The core tasks of the job, from the US Labor Department's O*NET profile (15-2051.00). Each verdict links to its evidence.
| Task | Verdict | Tools doing it | Why |
|---|---|---|---|
| Create graphs, charts and other visuals that show the results of an analysis | AI does this today | ChatGPT, Copilot in Power BI | ChatGPT turns an uploaded spreadsheet into bar, line and scatter charts, and Copilot in Power BI builds a whole report page from one sentence, so your job shifts to checking that the chart tells the truth. [3][5] |
| Clean and reshape raw data | AI helps, you check | ChatGPT, Data Science Agent in Colab | Google's Data Science Agent in Colab and ChatGPT both write and run the cleanup code for you, but they guess at messy headers and odd values, so you still decide what counts as an error. [3][4] |
| Analyze large data sets with statistical software | AI helps, you check | ChatGPT, Data Science Agent in Colab | ChatGPT runs Python in a Jupyter style sandbox for calculations and statistics, yet OpenAI warns it may pick a method you did not intend and tells you to review the code and assumptions. [3][4] |
| Write new functions or programs in Python, R or SQL to run analyses | AI helps, you check | ChatGPT, Data Science Agent in Colab | AI assistants draft working Python and SQL from a plain request, which removes boilerplate, but code that silently drops rows or leaks future data still has to be caught by you. [3][4] |
| Apply feature selection and feature engineering to prediction models | AI helps, you check | Data Science Agent in Colab | Colab's agent uses reasoning tools for feature engineering, but knowing which signals are real and which only exist in the training data comes from understanding the business. [4] |
| Test, validate and rebuild models so predictions hold up | AI helps, you check | ChatGPT | AI can write the validation code, but in the DSBench study the best agent closed only part of the gap to expert results on modeling tasks, so judging whether a model is good enough stays with you. [3][6] |
| Compare models using metrics such as loss functions or explained variance | AI helps, you check | ChatGPT | Generating a table of metrics is quick for AI, but choosing the metric that matches the real cost of a wrong prediction is a decision someone has to own. [3] |
| Find relationships, trends and factors that could skew results | AI helps, you check | ChatGPT, Copilot in Power BI | Copilot in Power BI answers questions about a model and points to the visuals behind each answer, which speeds up exploring, but spotting confounders and sampling bias needs domain sense. [3][5] |
| Present results of models and analysis to managers and other users | AI helps, you check | Copilot in Power BI | Power BI can write narrative summaries of report pages for you, but walking executives through uncertainty and answering their pushback is still a live conversation. [5] |
| Recommend data driven solutions to key stakeholders | Still human | None | A recommendation that changes budgets or staffing needs a person who understands the politics, the tradeoffs and who will be held to account if it is wrong. |
| Identify business problems that data analysis can solve | Still human | None | Picking the right question is the part AI cannot do from a spreadsheet, because the problem usually lives in meetings, incentives and things nobody wrote down. |
| Interview stakeholders to define the questions an analysis should answer | Still human | None | Getting a sales lead or a nurse manager to explain what they really need takes trust, follow up questions and reading the room. |
| Design surveys and sampling plans to collect new data | AI helps, you check | SurveyMonkey Build with AI | SurveyMonkey can draft a whole survey from one prompt, but Nielsen Norman Group found AI drafts often have unbalanced answer scales and too much respondent burden, so you still decide who to sample and pilot test the wording. [10][11] |
The job market for data scientists
Source: Data Scientists. BLS expects this job to grow 35% over 2025 to 2035.
Where data scientists can move next
GROWING · $140,300
Computer and information research scientist
These are the people who invent and improve the algorithms, including AI systems, so the work grows as AI spreads. Your math, coding and experiment skills carry over, though it is a small field of about 38,600 jobs. [7]
Median pay $140,300 · outlook +22%GROWING · $130,000
Actuary
Actuaries sign off on risk estimates for insurers and pension plans after passing a long series of professional exams, and that credential plus personal accountability is hard to automate. Your statistics and modeling background is a head start on the exams. [9]
Median pay $130,000 · outlook +9%GROWING · $88,940
Operations research analyst
Operations research turns models into concrete decisions about schedules, supply chains and staffing, which keeps you close to the people who act on the answer. Optimization and simulation skills transfer directly, though median pay is lower. [8]
Median pay $88,940 · outlook +12%AI tools changing this job
Our honest reviews of the tools named above.
Top picks: The Best AI Coding Assistant · The Best AI Chatbot
DATA SCIENTIST AI PROOF KIT
Want the step by step plan for data scientists?
We are writing a data scientist version of our AI Proof Kit: the tools to learn, how to set them up, a raise script and the safest next roles. We write kits in the order people ask for them.
Want to see what a kit looks like? See the bookkeeper kit.
Quick questions
Will AI replace data scientists in the next 10 years?
The official forecast points the other way. BLS projects data scientist employment to grow 35% from 2025 to 2035, with about 24,800 openings a year, and says data scientists will be needed to help businesses use AI based systems [2]. What will change is the daily work, as more cleaning, coding and charting is handed to AI tools [3] [4].
Can AI do data analysis without a data scientist?
For simple questions on a clean file, often yes, but it is unreliable on real projects. In the DSBench benchmark the best agent solved about 34% of realistic data analysis tasks [6]. OpenAI's own help page says ChatGPT may choose a method or chart you did not intend and tells users to review the generated code, outputs and assumptions [3].
What AI tools do data scientists use?
ChatGPT's data analysis feature reads spreadsheets and runs Python in a sandbox to calculate and chart results [3]. Google's free Data Science Agent in Colab helps clean data, engineer features and write SQL on uploaded CSV or JSON files under 1GB [4]. Copilot in Power BI drafts report pages, summaries and DAX queries [5].
Is data science still a good career with AI?
On the numbers, yes. The median data scientist earned $120,230 in May 2025, and growth is projected to be much faster than average [2]. Most jobs ask for a bachelor's degree in math, statistics or computer science, and some employers prefer a master's or PhD [2].
Which data science skills are safest from AI?
The tasks O*NET lists around people and judgment, such as interviewing stakeholders, framing business problems and recommending solutions, are the hardest to hand to software [1]. Charting and routine cleanup are the most exposed, since tools like ChatGPT and Power BI Copilot already do them from a short prompt [3] [5]. Even survey design is now drafted by AI, though experts say the drafts still need a trained reviewer [11].
Sources
Checked 2026-10-10. Verdicts are Leapscope's reading of these sources. We did not test these tools hands on for this page.
- Data Scientists (15-2051.00), O*NET OnLine
- Data Scientists, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics
- Data analysis with ChatGPT, OpenAI Help Center
- Google upgrades Colab with an AI agent tool, TechCrunch
- Use Copilot with Power BI reports and semantic models, Microsoft Learn
- DSBench: How Far Are Data Science Agents from Becoming Data Science Experts?, arXiv (ICLR 2025)
- Computer and Information Research Scientists, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics
- Operations Research Analysts, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics
- Actuaries, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics
- Build Surveys with AI, SurveyMonkey Help Center
- AI Can Help with Survey Writing, But It Still Requires Human Expertise, Nielsen Norman Group
More jobs: Accountants · Actuaries · Administrative assistants · Architects · Bank tellers · Bookkeepers · Cashiers · Claims adjusters · Computer support specialists · Copywriters · Court reporters · Customer service representatives · Data entry clerks · Electricians · Financial analysts · Graphic designers · Hr specialists · Insurance underwriters · Journalists · Lawyers · Legal secretaries · Loan officers · Management consultants · Market research analysts · Marketing managers · Medical coders · Medical transcriptionists · Paralegals · Payroll clerks · Pharmacists · Photographers · Project managers · Proofreaders · Radiologists · Real estate agents · Receptionists · Registered nurses · Retail salespeople · Sales representatives · Software developers · Tax preparers · Teachers · Technical writers · Telemarketers · Translators · Travel agents · Truck drivers · Video editors · Web developers