contributed by Jessica Matthews, AI and data science consultant and former VP of Global Product Management at Gartner
Is an AI Degree Worth It, or Is CS, Math, or Stats a Better Foundation?
According to Programs.com, there are now 450 AI master’s programs in the U.S., including AI concentrations within computer science and engineering degrees.
My perspective is that only the PhD degree helps you enter the AI field without prior experience with artifical intelligence. The only exceptions might be if you have a CS undergrad and a master’s in math or stats, or vice versa.
What a Berkeley Master’s Did and Didn’t Provide
I believe that for several reasons. Although I have an MS from a prestigious university, UC Berkeley, and a 10+ year background in research analytics, it is generally not enough. If you have work experience with data, that doesn’t count unless you have the data science/ML title. Once you have a master’s degree, you are in competition with PhDs.
An Example From Drug Discovery
As an example, I am interested in using AI for drug discovery. There is a master’s program at UCSF that focuses on data science for drug discovery specifically. However, after reviewing the job postings for the roles that graduates target, they all require a PhD. So more in-depth technical knowledge with less relevant domain knowledge is the only path.
Although much is said about the value of domain knowledge, in my experience, it is considered a nice-to-have in the hiring decision, where technical skills remain the differentiator.
How Has Generative AI Changed What Employers Want in New Hires?
Why Entry-Level Is Harder Now
What’s really hard right now for entry-level employees is that exploratory data analysis and data visualization can be done pretty effectively using LLMs and no technical skill. If you are working on something critical, you want a manual review, and if the datasets are very large, you need data engineering skills to complete the ETL to view the data, particularly in repeatable ways.
Beyond that, data skills like data governance and building semantic layers are valuable but really difficult to attain early in your career. Data engineering and developer roles are the most common starting point.
A 2025 Stanford Digital Economy Lab study found that early-career workers (ages 22 to 25) in the most AI-exposed occupations saw a 13% relative decline in employment after generative AI took off.
SQL Still Matters
The market still requires a high hiring bar for SQL skills, although SQL queries are easily automated. The inconsistent role definitions that have plagued the field for years continue. For example, data science roles often require ML model experience, productionizing models, and assisting with AI automation, three wildly different skill sets. New tools get tacked on (n8n, Make, Airtable, dbt), but SQL commands the same allegiance as it did in the 70s.
Are Degree Programs Keeping Up?
I question the move towards offering undergraduate degrees in AI and data science. The field requires a strong foundation in CS, stats, and AI toolsets that would be hard to cover meaningfully along with survey course requirements.
That said, if these programs are viewed as prerequisites to other jobs, such as data product managers, project managers, or PhD programs, and not core ML roles, they might make sense.
If I Were Starting Over Today, Would I Get an AI Degree?
If I were starting over today, I would look for project work on Upwork and see how far I could go solving business problems by learning commercial tools. Once I got to problems that require model development, I would begin to dig into foundational knowledge. The commercial tools have come so far that I fear we are paying to learn how to build calculators while people are being paid to operate checkout systems.
Alternatively, I would commit to the PhD path from the start, recognizing that an MS will only get you into competition with a large pool of PhDs. Although conventional wisdom says that business acumen and domain knowledge are critical, the labor market does not agree.