Artificial intelligence does not have to eliminate a profession before it changes education. Students make decisions years in advance based on what they believe the future job market will look like. As those expectations change, AI could begin reshaping majors, skills, degrees and career pathways long before its full impact on employment becomes clear.
THE SIGNAL
Technology usually changes the workforce first and education second.
A new technology arrives.
Companies adopt it.
Jobs change.
Schools eventually update what they teach.
Artificial intelligence may be disrupting that sequence.
Students don’t have to wait until AI actually eliminates jobs before reacting to the possibility that it might change them.
They are choosing majors today for careers they expect to enter four, six or even ten years from now.
That means expectations about AI can change education before AI completely changes employment.
Computer science may be providing one of the first visible examples.
For nearly two decades, computer science became one of American higher education’s biggest growth stories.
The logic was powerful:
Software was spreading everywhere.
Technology companies were expanding.
Developers were highly paid.
Learning to code appeared to provide one of the clearest pathways from college into a high-growth career.
That trend has now changed direction.
National Student Clearinghouse data shows undergraduate Computer and Information Sciences enrollment falling 8.4% at four-year institutions, 9.3% at primarily associate-degree granting baccalaureate institutions and 11.2% at two-year colleges in spring 2026.
That happened while overall undergraduate enrollment increased 1.3%. (National Student Clearinghouse)
Computer science also recorded the lowest second-year persistence rate among the ten largest bachelor’s-degree fields: 85%, compared with 93.1% for engineering. (National Student Clearinghouse)
At the same time, AI-related education is growing.
Stanford’s 2026 AI Index reports that computer science enrollment at U.S. four-year universities fell 11% between 2024 and 2025 while master’s graduates in AI software-related fields increased 17% from 2023 to 2024. (Stanford HAI)
Students may not be abandoning technology.
They may be repositioning themselves inside it.
WHAT THE MARKET IS MISSING
The obvious question is:
Is AI reducing computer science enrollment?
We don’t know.
And we should not claim that it is.
Computer science enrollment could be affected by many forces: the weaker entry-level technology market, the extraordinary growth of the major during the previous decade, changing student demographics, competition from data science and engineering programs, or broader economic conditions.
The more important signal is different.
Students are already incorporating AI into career decisions whether AI ultimately causes the job losses they fear or not.
Gallup and the Lumina Foundation found that 42% of bachelor’s-degree students had given at least a fair amount of thought to changing their major because of AI. Among associate-degree students, the number was 56%. (Gallup.com)
Another 2026 survey found 42% of college-eligible students said AI would influence which career they pursue, while 10% said they had already changed their planned major because of AI concerns. (Inside Higher Ed)
A CNBC/SurveyMonkey survey similarly found that four in ten students had considered changing their field of study because of AI, while roughly two-thirds believed entry-level workers were particularly vulnerable. (SurveyMonkey)
That means something important is already happening.
AI expectations are becoming an economic force of their own.
A student considering computer science does not need proof that software engineering will disappear.
They merely need to believe that:
- entry-level jobs may become harder to obtain,
- routine coding may become less valuable,
- employers may expect AI skills on day one,
- or another field may provide greater long-term security.
Once enough students make decisions based on those expectations, the education system begins changing even if the ultimate forecast turns out to be wrong.
That is the deeper signal.
FIRST-ORDER EFFECTS
Students begin pricing AI risk into majors
Students have always considered employment prospects when choosing degrees.
AI adds another variable:
automation exposure.
A prospective student may increasingly ask:
Will this profession still need as many entry-level workers when I graduate?
Which parts of this job will AI perform?
Will employers still train junior employees if AI handles junior work?
Will this degree still justify its cost?
Those questions can influence behavior long before anyone knows the definitive answers.
One EAB survey found a student who had moved from computer science toward electrical and computer engineering after observing AI’s impact on entry-level technology work. Other students reported reconsidering entirely different professions. (Inside Higher Ed)
That does not prove those students are correctly forecasting the future.
It proves that AI expectations are entering educational decision-making now.
Students move toward perceived complements to AI
Computer science enrollment may be cooling, but that doesn’t mean students are fleeing technical subjects.
National Student Clearinghouse data shows engineering added approximately 41,000 undergraduate students in spring 2026. Health professions added roughly 61,000, with four-year health enrollment surpassing one million students. (National Student Clearinghouse)
Meanwhile, AI-specific programs are expanding.
Northeastern Illinois University, for example, launched Illinois’ first public-university undergraduate AI major for fall 2026. (Axios)
This suggests the educational response may not simply be:
Avoid technology.
It may increasingly become:
Choose the part of technology that appears positioned to work with AI rather than compete against it.
Entry-level employment becomes part of the calculation
Students aren’t making these decisions in a vacuum.
The labor market is sending mixed but meaningful signals.
Stanford’s Digital Economy Lab reported in August 2026 that employment among workers ages 22 to 25 in highly AI-exposed occupations stands roughly 19% below where it would be if it had kept pace with similarly aged workers in less-exposed occupations.
The researchers found that the adjustment appears to be occurring primarily through reduced hiring of young workers rather than increased layoffs.
They also explicitly caution that these are descriptive patterns rather than definitive causal estimates of AI’s impact. (Stanford Digital Economy Lab)
That distinction is critical.
We are not seeing evidence that AI has destroyed the overall labor market.
We may be seeing early pressure at the entrance to certain professions.
For students, the entrance is what matters most.
SECOND-ORDER EFFECTS
This is where the implications become much larger than computer science.
Education could become a leading indicator of labor-market expectations
Economists traditionally watch employment, wages, job openings and productivity to understand technological disruption.
AI may create another useful indicator:
What are students choosing to study?
A falling major does not prove a profession is disappearing.
But sudden changes in enrollment can reveal how younger generations perceive the future.
Students are effectively making long-term bets with their time and tuition.
If thousands of people begin moving away from one field and toward another, that movement can tell us something about future expectations before employment statistics fully catch up.
Education may therefore become an early-warning system for technological change.
Perception can eventually change reality
There is also a feedback loop.
Suppose students believe software engineering is becoming less attractive.
Fewer students study computer science.
Several years later, fewer traditional computer science graduates enter the workforce.
Companies respond by hiring from adjacent disciplines, increasing automation, recruiting internationally, or valuing alternative credentials.
Universities respond by shrinking traditional programs and expanding AI, robotics, cybersecurity or interdisciplinary degrees.
Employers then redesign job descriptions around the new talent supply.
What began as an expectation can eventually help create the labor market students originally expected.
That is why perceptions matter.
They don’t merely predict markets.
Sometimes they influence them.
The traditional four-year degree faces a timing problem
A bachelor’s degree is a slow product operating inside an increasingly fast technology cycle.
A student entering college in fall 2026 may graduate in 2030.
An eighth grader choosing high-school coursework today may not enter the professional workforce until the mid-2030s.
AI capabilities can change dramatically within months.
That creates a structural mismatch:
Students must make long-term educational bets about a technology changing on a short-term cycle.
Universities therefore face an increasingly difficult challenge.
They cannot redesign curricula every time a new model launches.
But they also cannot assume that a curriculum designed around the labor market of 2023 will automatically prepare a student graduating in 2030.
Majors may become less important than skill combinations
AI could also weaken the traditional idea that a single major maps cleanly onto a single profession.
The emerging advantage may come from combinations:
Computer science + business
Medicine + AI literacy
Finance + data
Engineering + robotics
Law + AI governance
Psychology + human-computer interaction
Communications + AI verification
Students may increasingly ask not:
What should I major in?
but:
What combination of human expertise and AI capability will remain valuable?
That could push universities toward more modular education, certificates, minors, interdisciplinary programs and continuing education.
Career planning could become continuous
The old model assumed that education happened first and work came afterward.
AI may weaken that boundary.
If professions continue changing rapidly, workers may need repeated education throughout their careers.
The degree becomes the foundation.
Learning continues indefinitely.
Stanford’s 2026 AI Index already finds that people are acquiring AI skills outside formal education while putting those skills on resumes. (Stanford HAI)
The future education market may therefore shift from:
Learn → graduate → work
toward:
Learn → work → relearn → work → adapt → relearn again.
WINNERS
Universities that adapt without chasing hype
The strongest institutions will not simply rename every course “AI.”
They will identify which foundational skills remain durable while updating the methods through which students apply them.
Computer science students still need algorithms, systems thinking, architecture and security.
Doctors still need medicine.
Lawyers still need law.
Accountants still need accounting.
AI changes the tools.
It does not eliminate the need for underlying knowledge.
The winning institutions may therefore teach both:
the durable foundation and the changing interface.
Students who build adaptable skill portfolios
Students who prepare for one narrowly defined job may face more risk than students who develop transferable capabilities.
Technical literacy.
Critical thinking.
Communication.
Domain expertise.
AI literacy.
Judgment.
Problem definition.
Verification.
Collaboration.
Those skills can travel across tools and industries.
The advantage may increasingly belong to students who can change with the profession rather than merely train for its current version.
AI-native educational programs
Schools capable of integrating AI into actual disciplines rather than treating it as a standalone novelty could gain demand.
An AI-literate nursing program.
AI-assisted engineering.
AI-enabled financial analysis.
AI-supported scientific research.
AI governance inside law.
Those combinations may become more valuable than generic AI instruction because they connect the technology to actual professional judgment.
Fields perceived as complementary to AI
Healthcare, certain engineering specialties, skilled technical work and professions requiring physical interaction or significant human accountability may become more attractive if students view them as relatively resistant to pure software automation.
That does not mean those occupations are immune to AI.
It means students may increasingly distinguish between work they believe AI will replace and work they believe AI will amplify.
LOSERS
Programs that teach yesterday’s job
The greatest risk may not belong to any particular academic discipline.
It belongs to programs whose curriculum remains tied to tasks employers increasingly automate.
If a four-year degree prepares students primarily to perform entry-level tasks that AI can handle by graduation, the degree’s value proposition becomes difficult to defend.
That applies beyond computer science.
Business.
Accounting.
Marketing.
Media.
Law.
Finance.
Design.
Research.
Any field can face the same problem.
Students who overreact to uncertain forecasts
There is another side to this.
Students can make bad bets too.
Today’s predictions about which jobs AI will replace could prove wrong.
AI may complement many professions rather than eliminate them.
New occupations may appear.
Fields students abandon could eventually face talent shortages.
Computer science itself may become more important, not less, if AI increases the total amount of software society produces.
Students who abandon foundational technical education solely because AI can generate code could discover that the people who understand computing deeply become even more valuable.
That is why AIMainstream should not frame enrollment declines as proof that students are making the correct decision.
Behavior is changing. Whether the behavior is correct remains unknown.
Universities that move too slowly
Higher education operates through committees, accreditation, faculty structures and multi-year program cycles.
Technology companies can change workflows in weeks.
That speed difference creates risk.
Institutions that take years to determine whether AI belongs in a curriculum may eventually be teaching students tools and processes employers have already replaced.
Universities that move too fast
The opposite mistake is equally dangerous.
Schools can chase every technology trend and weaken foundational education.
Students trained around today’s AI interface rather than durable knowledge may discover that the interface is obsolete before graduation.
The challenge is not teaching today’s hottest tool.
It is identifying which capabilities remain valuable when the tools change.
WHAT HAPPENS NEXT
The computer science numbers deserve close attention, but the next signal may come from somewhere else.
Watch major switching, not simply enrollment.
Gallup’s finding that 42% of bachelor’s students have already considered changing their major because of AI suggests the potential movement is much larger than current enrollment statistics reveal. (Gallup.com)
Watch which fields gain students.
If traditional computer science continues declining while AI, engineering, healthcare, robotics, cybersecurity and other fields expand, that would suggest students are not rejecting technology but reallocating toward areas they believe have stronger future value.
Watch entry-level hiring.
Stanford’s latest labor-market research finds no widespread AI-driven employment collapse, but it does find increasing weakness among young workers in exposed occupations. If that pattern expands, students will have stronger reasons to adjust educational decisions. (Stanford Digital Economy Lab)
Watch new majors and certificates.
Universities will likely respond to student demand by introducing more programs combining established disciplines with AI.
Watch parents and high-school students.
The effects may eventually begin before college.
EAB found that 46% of surveyed high-school students were already using AI during the college-search process in early 2026, meaning AI is increasingly influencing not only what students learn but how they choose where to learn it. (EAB)
And watch whether employers begin moving away from degree titles toward demonstrable skills.
If AI repeatedly changes the tasks within jobs, hiring managers may care increasingly about whether candidates can adapt rather than whether their diploma perfectly matches a job description written years earlier.
The central question will not be whether computer science enrollment rebounds next semester.
It will be whether the relationship between education and employment expectations has fundamentally accelerated.
Students used to look at the jobs available today and choose education accordingly.
Increasingly, they may be looking at what AI could do tomorrow.
BOTTOM LINE
Artificial intelligence does not need to replace millions of workers before it changes society.
It only needs to change what people believe the future will require.
A student selecting a major in 2026 is not choosing for the labor market of 2026.
They are choosing for 2030, 2035 and potentially decades beyond that.
That makes every major decision partly a forecast.
Computer science provides an important early signal.
After nearly two decades of extraordinary growth, enrollment is now declining while AI systems are becoming much more capable at software development.
We cannot conclude that AI caused the decline.
But we also cannot ignore the broader evidence that students are actively reconsidering majors and careers because of AI.
That means the educational effects of artificial intelligence may arrive before its final labor-market effects can even be measured.
The sequence could be:
AI capabilities change.
Expectations change.
Students change behavior.
Universities change programs.
The talent pipeline changes.
Then the workforce changes.
If that pattern holds, the first major workforce disruption from AI may not appear in a layoff announcement.
It may appear several years earlier in a college enrollment office.
