Home › Computer & Math › Data Scientists

How a data scientist can use AI, task by task

O*NET 15-2051.00: Data ScientistsMedian pay $120,230/year (BLS, May 2025)16 tasksJob Zone 4: Considerable preparation

Develop and implement a set of techniques or analytics applications to transform raw data into meaningful information using data-oriented programming languages and visualization software. Apply data mining, data modeling, natural language processing, and machine learning to extract and analyze information from large structured and unstructured datasets. Visualize, interpret, and report data findings. May create dynamic data reports.

Your job title Type your own title and every prompt on this page updates to match. Also called: —.

Where AI fits in this job

AI-ready 29%AI-assist 62%Stays human 8%

Share of the job's tasks, weighted by how important and how frequent each task is (O*NET ratings). 5 AI-ready, 10 AI-assist and 1 human-led tasks out of 16. This measures where AI can help, not whether the job will disappear.

Information & paperwork

Thinking & planning

People & communication

Leading & teaching

Hands-on & physical

Shape of the job: importance of each kind of work activity (O*NET, 1 to 5).

🗓 Start here on Monday

The most frequent, most important tasks where AI can help right away.

Analyze, manipulate, or process large sets of data using statistical software.Crunch the numbers · Daily
Create graphs, charts, or other visualizations to convey the results of data analysis using specialized software.Create & design · Daily
Compare models using statistical performance metrics, such as loss functions or proportion of explained variance.Crunch the numbers · Weekly

💪 Lean into these

Core tasks that stay human. These skills get more valuable as the routine work gets automated.

Clean and manipulate raw data using statistical software.Hands-on work · Daily

Skills to build & where you could move ↓

AI can draft these 5

AI produces a usable first pass. You review it, fix it and own it.

Analyze, manipulate, or process large sets of data using statistical software.

Daily

AI finds trends and writes the spreadsheet formulas.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 50% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

Paste a spreadsheet or table and ask AI to find trends, outliers and totals, or to write the formula you need. It's like having an analyst on call.

Ready-to-use prompt

I work as a data scientist. Part of my job is to analyze, manipulate, or process large sets of data using statistical software.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Decision to make: [approve, change, invest, staff]
Reports: [paste key figures or summaries]

Summarize what the reports show, the options, the trade-offs, and what you'd want to know before deciding. Show calculations.

10-minute exercise: Turn reports into a decision

Step 1Take the reports in front of you this week.

Step 2Fill in [Decision to make] and [Reports] with the real details, then run the prompt.

Step 3Make the decision yourself, then note whether the analysis changed your mind.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 4.3/5 · core task · O*NET work activity: Analyze data to inform operational decisions or activities..

Identify solutions to business problems, such as budgeting, staffing, and marketing decisions, using the results of data analysis.

Weekly

AI finds trends and writes the spreadsheet formulas.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 60% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

Paste a spreadsheet or table and ask AI to find trends, outliers and totals, or to write the formula you need. It's like having an analyst on call.

Ready-to-use prompt

I work as a data scientist. Part of my job is to identify solutions to business problems, such as budgeting, staffing, and marketing decisions, using the results of data analysis.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Problem: [delays, errors, defects, protocol issues]
Data: [paste logs, indicators, printouts]

Find patterns that point to the cause, rank likely causes, and suggest checks and fixes. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Root-cause from data

Step 1Pick a real example: the next time you identify solutions (this week).

Step 2Fill in [Problem] and [Data] with the real details, then run the prompt.

Step 3Run the first check it suggests and confirm the cause before fixing. Recompute every total yourself.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Double-check every number. AI can make arithmetic and rounding mistakes.

Importance 3.9/5 · core task · O*NET work activity: Analyze data to identify or resolve operational problems..

Write new functions or applications in programming languages to conduct analyses.

Weekly

AI writes, explains and tests code you review.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 59% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

AI coding assistants can write, explain, test and refactor code from a plain description. You review, test and own what ships.

Ready-to-use prompt

I work as a data scientist. Part of my job is to write new functions or applications in programming languages to conduct analyses.

Context: We use Microsoft Azure software and C; the work draws on mathematics and computers and electronics.
Language and context: [language, existing code]
Goal: [what it should do]
Constraints: [style, performance, security]

Propose an approach, then write the code with comments and tests. Propose the approach first, then the code with comments and tests, and point out security risks.

10-minute exercise: Write code

Step 1Pick a real example: the next time you write new functions or applications (this week).

Step 2Fill in [Language and context], [Goal] and [Constraints] with the real details, then run the prompt.

Step 3Run the tests yourself and read every line before merging.

Don't paste secrets, passwords or proprietary code into tools your employer hasn't approved.

Importance 3.6/5 · core task · O*NET work activity: Write computer programming code..

Read scientific articles, conference papers, or other sources of research to identify emerging analytic trends and technologies.

Monthly

AI is a patient tutor for new rules and techniques.

Show prompt & steps

AI is a patient personal tutor. It explains new rules, techniques and technology at your level, quizzes you and builds a study plan.

Ready-to-use prompt

I work as a data scientist. Part of my job is to read scientific articles, conference papers, or other sources of research to identify emerging analytic trends and technologies.

Context: The work draws on mathematics and computers and electronics.
Field: [data science, landscape design, engineering]
Topics: [new methods, products, trends]

Explain the key developments in 5 short sections with examples, quiz me, and list sources to read. I'll verify sources exist.

10-minute exercise: Technical knowledge refresh

Step 1Pick a real example: the next time you read scientific articles and conference papers.

Step 2Fill in [Field] and [Topics] with the real details, then run the prompt.

Step 3Read one source it lists in full.

AI's knowledge has a cutoff date. For new regulations and standards, always confirm against the official source.

Importance 3.4/5 · core task · O*NET work activity: Update technical knowledge..

Design surveys, opinion polls, or other instruments to collect data.

A few times a year

AI writes the first draft from your notes; you check every fact.

Show prompt & steps

Give AI your rough notes, numbers or last version and it will turn them into a clean, well-structured first draft. You check every fact and figure before it goes anywhere.

Ready-to-use prompt

I work as a data scientist. Part of my job is to design surveys, opinion polls, or other instruments to collect data.

Activity: [program, initiative, campaign]
Questions: [what success means]

Draft evaluation methods, data collection and analysis. Use plain language and headings where they help, and mark anything you had to guess with [CHECK].

10-minute exercise: Evaluation procedure

Step 1Pick a real example: the next time you design surveys and opinion polls.

Step 2Fill in [Activity] and [Questions] with the real details, then run the prompt.

Step 3Check you can actually collect the data.

AI will happily make up numbers, names and dates. Every fact in the draft needs checking against your source.

Importance 3.0/5 · O*NET work activity: Develop procedures to evaluate organizational activities..

AI can help you prep these 10

AI speeds up the prep, drafting or thinking. The core of the task is still yours.

Create graphs, charts, or other visualizations to convey the results of data analysis using specialized software.

Daily

AI generates options and references; taste and the final call are yours.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 42% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

Use AI to generate lots of options, references and rough mock-ups fast. The craft, the taste and the final call are what make it yours.

Ready-to-use prompt

I work as a data scientist. Part of my job is to create graphs, charts, or other visualizations to convey the results of data analysis using specialized software.

Context: The work draws on mathematics and computers and electronics.
Project: [web page, mock-up, storyboard]
Goals and audience: [describe]

Generate 5 concept directions with layouts and styles. Give 10 quite different options from safe to bold, then develop the best 2.

10-minute exercise: Visual concepts

Step 1Pick a real example: the next time you create graphs and charts (today).

Step 2Fill in [Project] and [Goals and audience] with the real details, then run the prompt.

Step 3Pick the idea that sparks something and develop it yourself for 5 minutes.

Check the usage rules for AI images and text in your field, and make sure the final work is original.

Importance 4.3/5 · core task · O*NET work activity: Prepare graphics or other visual representations of information.. Also involves: Prepare analytical reports..

Compare models using statistical performance metrics, such as loss functions or proportion of explained variance.

Weekly

AI does the arithmetic and suggests what to look at; the conclusions are yours.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 52% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

AI can do the arithmetic, build the spreadsheet formula and suggest what to look at. Deciding what the numbers mean for your situation is still you.

Ready-to-use prompt

I work as a data scientist. Part of my job is to compare models using statistical performance metrics, such as loss functions or proportion of explained variance.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Problem: [describe]
Data: [paste or describe]

Suggest the right method, assumptions, and code. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Apply a statistical method

Step 1Pick a real example: the next time you compare models (this week).

Step 2Fill in [Problem] and [Data] with the real details, then run the prompt.

Step 3Recompute the most important number yourself before you trust the rest.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 4.0/5 · core task · O*NET work activity: Apply mathematical principles or statistical approaches to solve problems in scientific or applied fields..

Test, validate, and reformulate models to ensure accurate prediction of outcomes of interest.

Weekly

AI does the arithmetic and suggests what to look at; the conclusions are yours.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 47% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

AI can do the arithmetic, build the spreadsheet formula and suggest what to look at. Deciding what the numbers mean for your situation is still you.

Ready-to-use prompt

I work as a data scientist. Part of my job is to test, validate, and reformulate models to ensure accurate prediction of outcomes of interest.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Problem: [describe]
Data or principles: [inputs]

Propose a model and solve it with code. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Mathematical model

Step 1Pick a real example: the next time you test, validate and reformulate models to ensure accurate (this week).

Step 2Fill in [Problem] and [Data or principles] with the real details, then run the prompt.

Step 3Recompute the most important number yourself before you trust the rest.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 4.2/5 · core task · O*NET work activity: Develop scientific or mathematical models..

Identify relationships and trends or any factors that could affect the results of research.

Weekly

AI does the arithmetic and suggests what to look at; the conclusions are yours.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 36% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

AI can do the arithmetic, build the spreadsheet formula and suggest what to look at. Deciding what the numbers mean for your situation is still you.

Ready-to-use prompt

I work as a data scientist. Part of my job is to identify relationships and trends or any factors that could affect the results of research.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Data: [paste table]
Question: [what relationships matter]

Analyze correlations and trends with code or formulas. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Find trends and relationships

Step 1Pick a real example: the next time you identify relationships and trends or any factors (this week).

Step 2Fill in [Data] and [Question] with the real details, then run the prompt.

Step 3Recompute the most important number yourself before you trust the rest.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 3.7/5 · core task · O*NET work activity: Analyze data to identify trends or relationships among variables..

Identify business problems or management objectives that can be addressed through data analysis.

Weekly

AI gets you up to speed; confirm findings with records or people.

Show prompt & steps

AI can get you up to speed fast and suggest where to look, but you need to confirm findings against real sources, records or people.

Ready-to-use prompt

I work as a data scientist. Part of my job is to identify business problems or management objectives that can be addressed through data analysis.

Context: We use SAS and The MathWorks MATLAB; the work draws on mathematics and computers and electronics.
Project or problem: [describe]
Objectives: [goals]

List resources, staff and funding needed. Give sources I can check and say clearly when you're unsure.

10-minute exercise: Resource needs

Step 1Pick a real example: the next time you identify business problems or management objectives (this week).

Step 2Fill in [Project or problem] and [Objectives] with the real details, then run the prompt.

Step 3Open two of the sources it gives and confirm the key claim before you use it.

AI can invent sources and citations. Open every link and confirm it says what the AI claims.

Importance 4.1/5 · core task · O*NET work activity: Select resources needed to accomplish tasks..

Deliver oral or written presentations of the results of mathematical modeling and data analysis to management or other end users.

Monthly

AI helps you prepare clearer wording and likely questions.

Show prompt & steps

Use AI to prepare: simpler wording, analogies, likely questions and a one-page handout. The conversation itself is yours.

Ready-to-use prompt

I work as a data scientist. Part of my job is to deliver oral or written presentations of the results of mathematical modeling and data analysis to management or other end users.

Context: The work draws on mathematics and computers and electronics.
Audience: [management, clients, meeting]
Results: [paste key findings and numbers]

Turn this into a short presentation: headline message, 3 key charts to make, plain-language explanations, and likely questions with answers.

10-minute exercise: Present research results

Step 1Pick a real example: the next time you deliver oral or written presentations of the results.

Step 2Fill in [Audience] and [Results] with the real details, then run the prompt.

Step 3Check every number on your slides against the analysis output.

Rules, prices and policies change. Make sure AI is working from your current official version.

Importance 4.2/5 · core task · O*NET work activity: Present research results to others..

Apply feature selection algorithms to models predicting outcomes of interest, such as sales, attrition, and healthcare use.

Weekly

AI does the arithmetic and suggests what to look at; the conclusions are yours.

Show prompt & steps

AI can do the arithmetic, build the spreadsheet formula and suggest what to look at. Deciding what the numbers mean for your situation is still you.

Ready-to-use prompt

I work as a data scientist. Part of my job is to apply feature selection algorithms to models predicting outcomes of interest, such as sales, attrition, and healthcare use.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Question: [price reasonableness, cost trends, profitability]
Data: [paste proposals, reports, history]

Analyze, compare to benchmarks and give a recommendation. Show formulas. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Business data analysis

Step 1Pick a real example: the next time you apply feature selection algorithms (this week).

Step 2Fill in [Question] and [Data] with the real details, then run the prompt.

Step 3Use it in your next negotiation or estimate.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 3.8/5 · core task · O*NET work activity: Analyze business or financial data..

Recommend data-driven solutions to key stakeholders.

Monthly

AI helps you prepare clearer wording and likely questions.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 30% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

Use AI to prepare: simpler wording, analogies, likely questions and a one-page handout. The conversation itself is yours.

Ready-to-use prompt

I work as a data scientist. Part of my job is to recommend data-driven solutions to key stakeholders.

Context: We use SAS and The MathWorks MATLAB; the work draws on mathematics and computers and electronics.
Question: [modeling, data]
Audience: [departments]

Explain approaches and pitfalls. Use plain language, one everyday analogy, and the questions people are most likely to ask with short answers.

10-minute exercise: Analytics advice

Step 1Pick a real example: the next time you recommend data-driven solutions.

Step 2Fill in [Question] and [Audience] with the real details, then run the prompt.

Step 3Check every answer is right for your workplace before you share it.

Rules, prices and policies change. Make sure AI is working from your current official version.

Importance 4.2/5 · core task · O*NET work activity: Advise others on analytical techniques..

Propose solutions in engineering, the sciences, and other fields using mathematical theories and techniques.

Monthly

AI does the arithmetic and suggests what to look at; the conclusions are yours.

📊 Real AI use (Anthropic Economic Index, Apr 2026 to May 2026): AI completed the task directly in 53% of those conversations; the rest were back-and-forth with the person.

Show prompt & steps

AI can do the arithmetic, build the spreadsheet formula and suggest what to look at. Deciding what the numbers mean for your situation is still you.

Ready-to-use prompt

I work as a data scientist. Part of my job is to propose solutions in engineering, the sciences, and other fields using mathematical theories and techniques.

Context: We use Tableau and Apache Spark; the work draws on mathematics and computers and electronics.
Problem: [describe]
Data: [paste or describe]

Suggest the right method, assumptions, and code. Show your working step by step and give spreadsheet formulas so I can reproduce the numbers.

10-minute exercise: Apply a statistical method

Step 1Pick a real example: the next time you propose solutions.

Step 2Fill in [Problem] and [Data] with the real details, then run the prompt.

Step 3Recompute the most important number yourself before you trust the rest.

AI makes arithmetic mistakes. Recompute anything that matters with a spreadsheet formula.

Importance 3.0/5 · core task · O*NET work activity: Apply mathematical principles or statistical approaches to solve problems in scientific or applied fields..

Apply sampling techniques to determine groups to be surveyed or use complete enumeration methods.

A few times a year

AI lays out options and timelines to react to.

Show prompt & steps

AI can lay out options, timelines and checklists to react to. Fitting it to your people, budget and situation is your call.

Ready-to-use prompt

I work as a data scientist. Part of my job is to apply sampling techniques to determine groups to be surveyed or use complete enumeration methods.

Context: We use Microsoft Excel; the work draws on mathematics and computers and electronics.
Problem: [describe]
Data: [type, size, structure]

Recommend methods with assumptions, pros and cons, and starter code. Lay it out as a table with steps, owner, timing and what could go wrong, then list the decisions I still need to make.

10-minute exercise: Choose an analysis method

Step 1Pick a real example: the next time you apply sampling techniques to determine groups to be.

Step 2Fill in [Problem] and [Data] with the real details, then run the prompt.

Step 3Cross out anything that won't work on the ground and ask AI to revise with your corrections.

AI doesn't know your people, your site or your history. Treat its plan as a starting point to react to.

Importance 3.0/5 · core task · O*NET work activity: Determine appropriate methods for data analysis..

These stay human 1

They need your hands, your presence or your accountable judgment. Lean into them: they get more valuable as routine work is automated.

📏 Measured in this job (O*NET work context): Face-to-face discussions: once a week or more but not every day

Clean and manipulate raw data using statistical software.

Daily

Physical, in-person work. AI helps with checklists and write-ups.

Show prep prompt & steps

Where AI still helps: AI helps around the job: decoding manuals and instructions, pre-job checklists, calculations and writing up what you did afterward.

Prep prompt

I work as a data scientist. Part of my job is to clean and manipulate raw data using statistical software.

Context: The work draws on mathematics and computers and electronics.
What's involved: [raw data: where, when, who's affected]

Make me a step-by-step checklist to clean and manipulate raw data, the details I should double-check before I start, and a 3-line template for noting it's done.

10-minute exercise: Checklist: raw data

Step 1Before you next clean and manipulate raw data, fill in [What's involved] with the real details.

Step 2Strike any step that doesn't match how your workplace does it, and add what only experience teaches.

Step 3Afterward, note it with the 3-line template and one thing you'd do differently next time.

📊 People already use AI around this task (Anthropic Economic Index, Apr 2026 to May 2026). The task itself stays human-led.

Importance 4.0/5 · core task · O*NET work activity: Prepare data for analysis..

Get the full prompt pack for Data Scientists

Every task, prompt and 10-minute exercise on this page in one PDF, formatted to print or keep on your desktop. Free.

Your next moves

Skills to build

From O*NET's ratings of how important each skill is in this job, matched to which part of the work it supports.

Stays human Deepen these

They power the human-led work, which is the part of the job AI can't take over.

Programmingimportance 3.8/5
Used in: Working with Computers

With AI Use these to work with AI

When AI does the first draft, these are the skills you use to brief it and check what comes back.

Critical Thinkingimportance 4.2/5
Used in: Analyzing Data or Information
Reading Comprehensionimportance 3.9/5
Used in: Analyzing Data or Information
Complex Problem Solvingimportance 3.8/5
Used in: Making Decisions and Solving Problems
Active Listeningimportance 3.5/5
Used in: Making Decisions and Solving Problems

Where you could move

Related jobs from O*NET with the closest skill profile, favouring ones where more of the work stays human. No step down in preparation.

Mathematicians10% human-led · Needs more preparation · $126,710/yr · usually doctoral degree
Skills to build: Science, Systems Analysis, Systems Evaluation
Data Warehousing Specialists0% human-led · Same preparation · $139,500/yr · usually bachelor's degree
Skills to build: Programming, Systems Evaluation, Systems Analysis
Statisticians3% human-led · Needs more preparation · $105,650/yr · usually master's degree
Skills to build: Science, Systems Evaluation, Systems Analysis
Financial Quantitative Analysts0% human-led · Needs more preparation · $81,100/yr · usually master's degree
Skills to build: Systems Evaluation, Systems Analysis

New tasks appearing in this job

O*NET has started tracking 1 new or updated task here: 0 AI-ready, 1 AI-assist, 0 stays human. They have no worker ratings yet, so they aren't counted in the bar above.

Interview stakeholders to identify questions or problems to address through data analysis.

New in O*NET

AI helps you prepare questions; the listening is yours.

Show prompt & steps

Use AI to prepare better questions and to organize your notes afterward. Listening to the person in front of you is yours.

Ready-to-use prompt

I work as a data scientist. Part of my job is to interview stakeholders to identify questions or problems to address through data analysis.

Context: The work draws on mathematics and computers and electronics.
Who I'm talking to: [users and stakeholders, without personal details]
What I need to find out about stakeholders: [goals]

Write 8 open questions in a sensible order, follow-up prompts for vague answers, and a note-taking template I can fill in during the conversation.

10-minute exercise: Question list: stakeholders

Step 1Write down the questions you normally ask users and stakeholders to interview stakeholders.

Step 2Put the person type in [Who I'm talking to] and your aims in [goals], and ask AI what you're missing.

Step 3Use one new question in your next real conversation and note what it surfaced.

Don't record people or paste their answers into AI without consent and your employer's approval.

New O*NET task (08/2026). O*NET hasn't linked it to work activities yet, so it was labelled by hand.

Getting into this job

O*NET has not surveyed education and training for this occupation yet.

Software in this job

In-demand tools from O*NET. Many now have AI built in, so check what yours can do before adding a new tool.

Alteryx softwareAmazon Elastic Compute Cloud EC2Amazon RedshiftAmazon Web Services AWS softwareApache AirflowApache CassandraApache HadoopApache HiveApache KafkaApache SparkAtlassian ConfluenceAtlassian JIRA