---
title: "Will AI Replace Data Science? The Data on 16 Tasks | Nexum"
description: "Will data science be replaced by AI? AI could help with 15 of 16 data scientist tasks, yet the government expects 35% more jobs by 2035."
url: https://nexumautomations.com/roles/data-scientist
updated: 2026-11-10
publisher: Nexum Automations, an AI consulting firm
---

> Nexum Automations is an AI consulting firm that builds what it recommends. It finds where AI pays off in a company, builds it on the tools the team already uses and stays until it is live and working.

Roles

# Will AI replace data science ?

Here are the 16 tasks of a data scientist, checked against what AI could do. We also add the government's numbers on jobs and pay.

Last reviewed November 10, 2026

The short answer

Not as a whole job. People who studied the 16 data scientist tasks found 15 where AI could help. AI alone could speed up 4 of those. Even so, the U.S. Bureau of Labor Statistics sees 35% more jobs coming by 2035.

Key points

* AI could help with 15 of 16 tasks
* AI alone could speed up 4 of those 15
* 1 task stays human: comparing models with statistics
* Data scientist jobs expected up 35% by 2035
* People already use AI for 1 of 16 tasks

Tasks

## The tasks, one by one

Each row is a real task of this job (15 of 16). The columns show if AI could do it faster, and if people already use AI for it.

| | Can AI speed it up? | Already done with AI? |
| ---------------------------------------------------------------------------------------------------------------------------------- | ------------------------ | --------------------- |
| Analyze, manipulate, or process large sets of data using statistical software. | Yes, with extra software | Not seen yet |
| Apply feature selection algorithms to models predicting outcomes of interest, such as sales, attrition, and healthcare use. | Yes, with extra software | Not seen yet |
| Apply sampling techniques to determine groups to be surveyed or use complete enumeration methods. | Yes, AI alone | Not seen yet |
| Clean and manipulate raw data using statistical software. | Yes, with extra software | Not seen yet |
| Compare models using statistical performance metrics, such as loss functions or proportion of explained variance. | No | Not seen yet |
| Create graphs, charts, or other visualizations to convey the results of data analysis using specialized software. | Yes, with extra software | Not seen yet |
| Deliver oral or written presentations of the results of mathematical modeling and data analysis to management or other end users. | Yes, with extra software | Not seen yet |
| Design surveys, opinion polls, or other instruments to collect data. | Yes, with extra software | Not seen yet |
| Identify business problems or management objectives that can be addressed through data analysis. | Yes, with extra software | Not seen yet |
| Identify relationships and trends or any factors that could affect the results of research. | Yes, with extra software | Yes, seen in use |
| Identify solutions to business problems, such as budgeting, staffing, and marketing decisions, using the results of data analysis. | Yes, with extra software | Not seen yet |
| Propose solutions in engineering, the sciences, and other fields using mathematical theories and techniques. | Yes, AI alone | Not seen yet |
| Read scientific articles, conference papers, or other sources of research to identify emerging analytic trends and technologies. | Yes, with extra software | Not seen yet |
| Recommend data-driven solutions to key stakeholders. | Yes, AI alone | Not seen yet |
| Test, validate, and reformulate models to ensure accurate prediction of outcomes of interest. | Yes, with extra software | Not seen yet |

Tasks come from O\*NET, the U.S. government list of what each job does (code 15-2051.00). "Can AI speed it up?" comes from a 2023 research study ("GPTs are GPTs"). People who studied each task judged if an AI tool like ChatGPT could do it at least twice as fast. "Yes, AI alone" means the AI tool could do it by itself. "Yes, with extra software" means it would need other programs built around it. "Already done with AI?" comes from Anthropic data on how people use its AI at work. AI that could help with a task does not mean the job goes away.

Our data

## The job in numbers

We counted all 16 tasks of this job and added the government numbers on pay and jobs.

| | Number |
| -------------------------------------------------- | -------- |
| Tasks in this job (government list) | 16 |
| Tasks AI could speed up, say people | 15 of 16 |
| Tasks AI could speed up on its own | 4 of 16 |
| Tasks people already use AI for | 1 of 16 |
| Share of the work already done with AI (Anthropic) | 46.1% |
| Typical yearly pay, data scientists (May 2025) | $120,230 |
| Jobs in 2025, data scientists | 275,600 |
| Expected change in jobs by 2035, data scientists | +35% |

Task counts by Nexum from O\*NET 15-2051.00 and the 2023 study by Eloundou, Manning, Mishkin and Rock; the share already done with AI comes from the Anthropic Economic Index. Pay, jobs and the 2035 outlook come from the U.S. Bureau of Labor Statistics (https://www.bls.gov/ooh/math/data-scientists.htm). Typical pay is the middle point: half of workers earn more, half earn less.

## What stays with data scientists

Stays with people

* Picking the right model score
* Defending a model choice
* Knowing the company and its limits

AI could speed up

* Cleaning raw data
* Drafting charts and visuals
* Writing up results for managers

Only one of the 16 tasks was marked as staying with people. It is weighing models against each other with math scores, such as how far off their guesses are.

Someone has to pick the score that fits the business question and stand behind that choice when the numbers disagree.

Many other tasks keep a human core too. Spotting which business problem a team should tackle depends on knowing the company, its customers and its limits.

## Three paths for data science jobs

* AI helps out
Scientists let AI clean and draft, then focus on judgment
* Work shifts
More time choosing problems and checking model output
* Fewer people needed
Routine analysis needs fewer people where firms automate heavily

These are three ways the same task list could play out. In the first, data scientists let AI clean and draft, then spend the saved time on judgment.

In the second, the work moves toward choosing the problem and checking what models produce. In the third, routine analysis needs fewer people at firms that automate a lot.

Which one fits depends on the company, its data and how far it trusts model output without a person checking.

## Jobs and pay for data scientists

* $120,230
Typical pay a year, May 2025
[U.S. Bureau of Labor Statistics](https://www.bls.gov/ooh/math/data-scientists.htm)
* 275,600
Data scientist jobs in 2025
[U.S. Bureau of Labor Statistics](https://www.bls.gov/ooh/math/data-scientists.htm)
* 35%
Expected job growth, 2025 to 2035
[U.S. Bureau of Labor Statistics](https://www.bls.gov/ooh/math/data-scientists.htm)

Half of data scientists earn more than $120,230 a year and half earn less. That is the BLS figure for typical pay in May 2025.

It counts 275,600 data scientist jobs in 2025 and expects that number to grow 35% by 2035.

## Reading the numbers on this page

Four of the 16 tasks are ones AI could finish faster all by itself. The other 11 that AI could help with need extra software.

One task stands out: comparing models by how well they perform. People said AI can't help much there.

People already use AI for one task only: finding relationships and trends that could affect research results. So real chat use touches a small corner of the list. See [the method](https://nexumautomations.com/roles/method).

## If you run a business in this field

A data team spends hours on the work around the models. That means pulling data from several systems, building the weekly report pack and chasing people for answers.

That routine work can be automated while analysts keep choosing models and framing problems. See [how to automate data entry](https://nexumautomations.com/how-to/automate-data-entry) for a place to start.

A related read is [will AI replace data analysts](https://nexumautomations.com/guides/will-ai-replace-data-analysts), about the neighboring job.

FAQ

## Data scientists and AI questions

* ### Will data science be replaced by AI?
Partly, not all of it. AI could help with 15 of the 16 tasks, but AI alone could speed up only 4\. The BLS expects data scientist jobs to grow 35% by 2035.
* ### How will AI affect data science jobs?
AI is most likely to speed up cleaning data, drawing charts and writing drafts. The job could get help, shift toward judgment, or need fewer people where firms automate a lot.
* ### What does the data say about data scientist demand in future?
Demand keeps rising, with the BLS expecting 35% growth by 2035 from about 275,600 jobs in 2025\. Typical pay for a data scientist came to $120,230 in May 2025.
* ### Which data scientist tasks could AI help with?
Fifteen of 16, from cleaning raw data and building charts to presenting results. AI alone could speed up four, such as recommending data-driven solutions to managers. Comparing models on performance scores is the one task that stays with people.
* ### Is data science being replaced by AI in practice?
People already use AI for just 1 of the 16 tasks, finding relationships and trends in data. This data shows a direction, not every workplace. See [the method](https://nexumautomations.com/roles/method).

---

Source: https://nexumautomations.com/roles/data-scientist

Free AI diagnosis (30-minute call): https://nexumautomations.com/#book - Services: https://nexumautomations.com/services - Guides: https://nexumautomations.com/guides
