Building or testing AI systems for a living usually involves writing code, while using or overseeing them often does not. The U.S. Bureau of Labor Statistics states that data scientists write code, and it reports that computer and information research scientists, who work on algorithms, typically need at least a master's degree. Assessing the risk of an automated system, such as completing Canada's Algorithmic Impact Assessment, is done through a questionnaire and a multi-disciplinary team.
Key takeaways
- The U.S. Bureau of Labor Statistics counted about 275,600 data scientist jobs in 2025 and projects 35 percent growth through 2035; the work includes writing code.
- The federal O*NET database lists Programming among data scientists' transferable skills, and mathematics is the first essential skill on that page.
- Canada's Algorithmic Impact Assessment has 65 risk questions and 41 mitigation questions, and the government recommends a multi-disciplinary team to complete it.
- Whether coding is needed depends on the task: building a model, judging its impact, and managing its risk are different jobs.
Coding depends on the task, not on the label "AI"
The word "AI" covers very different kinds of work. Someone who trains a model, connects it to a database, or tests it against new data is doing software work. Someone who decides whether a government department may use an automated decision system, or who sets a company's rules for AI risk, is doing assessment and governance work. The table sorts four common tasks by whether the cited sources describe coding as part of them. It is a simplified orientation, and the entries marked "not stated" mean only that the source does not say, not that coding is forbidden.
| Task | Coding involved? | What the source says |
|---|---|---|
| Building and testing models for an employer | Yes | BLS: data scientists write code, and build and test algorithms and models |
| Designing new algorithms and computing tools | Yes, at depth | BLS: research scientists develop software systems and work with algorithms; a master's degree is typical |
| Rating the impact of an automated decision system | Not stated | Government of Canada: an online questionnaire, ideally completed by a multi-disciplinary team |
| Setting an organization's AI risk practices | Not stated | NIST: a voluntary framework covering the design, development, use, and evaluation of AI |
What U.S. labor data says about building roles
The Bureau of Labor Statistics profile of data scientists, last modified August 27, 2026, describes the job as identifying useful data, analyzing it, building and testing algorithms and models, and making business recommendations. It says the work includes writing code, and it advises students to learn data-oriented programming languages along with statistical and database software. High school students are told to take linear algebra, calculus, and probability and statistics.
The same profile reports about 275,600 jobs in 2025 and a projected rise to 371,000 by 2035, a gain of 95,400 and a growth rate of 35 percent, which the bureau calls much faster than the average for all occupations. The profile for computer and information research scientists, also modified in August 2026 (the latest editions as of October 2026), sets the typical entry requirement at a master's degree in computer science or a related field and lists advanced math among the relevant skills.
O*NET, the federal occupational database, points the same way. Its entry for data scientists lists Programming under transferable skills, and its software examples include languages such as Java, Go, and Scala alongside tools for statistics and cloud computing. The page marks some categories as partial, so the list is not exhaustive.
Roles that assess and govern AI
Not every AI-related job is a software job. The Government of Canada's Algorithmic Impact Assessment is a mandatory risk assessment tool that supports the Treasury Board's Directive on Automated Decision-Making. It asks 65 risk questions and 41 mitigation questions, sorts a system into one of four impact levels, and ties each level to requirements under the directive. The page recommends that departments complete it with a multi-disciplinary team, and it does not describe coding as a requirement.
In the United States, the National Institute of Standards and Technology released version 1.0 of its AI Risk Management Framework on January 26, 2023. The framework is voluntary and is meant to help build trustworthiness into the design, development, use, and evaluation of AI products, services, and systems. On July 26, 2024, NIST added a profile for generative AI. NIST says the framework is being revised as part of the White House AI Action Plan and gives no timeline. On April 7, 2026, it also published a concept note for a profile on trustworthy AI in critical infrastructure. Version 1.0 remains the latest full edition as of October 2026, so the current text should be checked before it is used in a policy document.
Matching a goal to a skill level
- If the goal is to build or train models professionally, expect programming and statistics to be part of the work, as the labor data describes.
- If the goal is to evaluate or buy an AI tool, the useful skills are reading performance claims, asking about training data, and understanding risk categories.
- If the goal is to write policy, the Canadian and NIST documents are written for organizations and teams rather than for individual software developers.
- If the goal is undecided, a short beginner programming course shows quickly whether the building side suits the learner.
FAQ
Can someone get a job in AI without knowing how to code?
Yes, in roles centered on assessment, policy, or project coordination, though the cited sources do not give a count of such jobs. Roles that build models, such as data scientist, involve writing code according to the Bureau of Labor Statistics.
Which programming language do AI roles require?
The sources checked do not name a single required language. O*NET lists several, including Java, Go, and Scala, and the Bureau of Labor Statistics refers to data-oriented languages in general.
Is advanced math needed as well as coding?
For research roles, the bureau lists advanced math among the relevant skills. For data scientists, it recommends linear algebra, calculus, and probability and statistics in high school.
Does a risk assessment of an AI system need a software developer?
The Canadian tool is a questionnaire and the government recommends a multi-disciplinary team, which can include technical staff. The page does not set a coding requirement.
Related reading
- Why AI systems still depend on code
- How much math AI work involves
- Whether AI counts as a computer
- How AI systems learn from new data
More in AI Basics.
Last verified: October 2026
Sources
- Data Scientists, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics, 2026
- Computer and Information Research Scientists, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics, 2026
- Data Scientists (15-2051.00), O*NET OnLine, U.S. Department of Labor, 2026
- Algorithmic Impact Assessment tool, Government of Canada, Treasury Board of Canada Secretariat, 2026
- AI Risk Management Framework, National Institute of Standards and Technology, 2023 and 2024