A Worst-Case Scenario: AI Destroys 80% of College

Worst-Case Scenario: AI Destroys 80% of College

Read This First

This is a stress test, not a prediction. We are asking a hard question on purpose: If Artificial Intelligence (AI) gets much better in the next 3–5 years, could most college degrees lose their value? What breaks first, who survives, and what should students, graduates, employers, and schools do?

Thesis: In a dark—but plausible—scenario, AI makes about 80% of degree programs less useful for jobs. The remaining 20% survive where hands-on work, labs, safety rules, required licenses, or cutting-edge research are essential (for example, medicine and engineering). This is not about “liking” or “disliking” college. It’s a drill to see where the model fails—and where it still wins.

Why This Is Not Crazy Talk

1) AI already passes hard tests.
GPT-4 has passed the Uniform Bar Exam with a score above the passing threshold, and research is actively debating how strong that performance is. Either way, AI is no longer just “cute chat.” It can handle complex prompts, write long answers, and “think” through structured tasks.

2) Big groups expect big job changes.
The World Economic Forum (WEF) says 23% of jobs will change within five years, with a net loss of ~14 million roles across the employers they surveyed. Goldman Sachs estimates the equivalent of 300 million jobs worldwide are exposed to automation by generative AI. IBM’s Institute for Business Value reports executives expect about 40% of workers will need to reskill within three years. (Reskilling = learning a new set of skills for a different role.)

3) Colleges are discounting more to fill seats.
Private nonprofit colleges set a new record tuition discount rate—56.3% for first-time undergrads in 2024–25. That means, on average, schools gave back 56 cents of every tuition dollar to fill classes. That is a red flag for pricing power.

4) Enrollment rebounded, but it’s still below 2019 totals.
Spring 2025 enrollment rose 3.2% year over year; undergrad counts are still ~2.4% below pre-pandemic. A bounce is not a fix.

5) Closures and mergers keep stacking up.
More than 80 public or nonprofit colleges have closed or merged since 2020, and trackers continue to add new cases.

What Would Have to Be True (Five Triggers)

For the “80% becomes irrelevant” story to happen, most of these must land together:

1) AI becomes a “doer,” not just a “talker.”
Models don’t just answer—they do work: click buttons, use apps, fill forms, write code, and run repeatable workflows. This “computer use” ability has moved from research demos to public betas (for example, Computer-Using Agent (CUA) from OpenAI’s Operator, and Computer Use from Anthropic). If this scales, many entry-level tasks go to AI by default.

2) Cost beats campus.
If AI tutoring, grading help, and project feedback become near-free and high-quality, a $30K semester is hard to justify—especially for lecture-and-essay classes.

3) Hiring shifts to proof over pedigree.
Employers continue to reduce degree requirements and rely more on work samples and skills tests. Indeed’s data shows fewer postings list degrees at all, with only ~18% requiring a four-year degree in early 2024. Harvard Business School (HBS) and the Burning Glass Institute call this shift the “Emerging Degree Reset.”

4) Schools stay financially fragile.
Discounts creep higher. Small, tuition-dependent colleges announce more teach-outs, mergers, or closures.

5) Policy doesn’t step in to prop up lecture-heavy majors.
If there’s no major public funding or policy move that protects non-lab programs, market pressure rules.

The Domino Chain (3–5 Years)

Year 0–1 (Right now to 12 months)

• AI tools become normal in offices. Many junior-level tasks (first drafts, short analyses, basic coding, slide creation, CRM updates) shift to AI. Companies talk more about skills-first hiring and run more short skills tests.
• Private colleges lift tuition discounts again to hit enrollment numbers.

Year 1–3

• More postings drop degree requirements or list them as “preferred,” not “required.” Work samples and task trials gain ground, even if not everywhere. HBS and Burning Glass see the reset spreading across many roles.
• Program cuts and closures/mergers speed up among tuition-dependent schools and low-demand majors.

Year 3–5

• A few AI-native programs appear, built around studios and clinics rather than lecture halls.
• Survivors cluster where hands-on reality matters: clinical fields, wet labs, engineering with physical prototyping, and elite research with heavy equipment.

Who Survives (Likely the “Protected 20%”)

• Medicine, nursing, allied health: clinical hours, patient safety, and required licenses aren’t going away.
• Engineering with labs and Accreditation Board for Engineering and Technology (ABET) limits: build, test, risk, and instrument handling are physical.
• Hard sciences with complex gear: fabrication labs, beamlines, biosafety facilities.
• Top-tier research: true frontier Research and Development (R&D), where new knowledge is created.

Who Is Most Exposed (Likely the “Commoditized 80%”)

• Lecture-centric, test-centric majors where outputs are texts, slides, simple dashboards, or baseline code—things AI drafts well.
• Coursework that AI tutors can practice at scale, without wet labs, clinicals, or licenses.

The Money Pressure You Can’t Ignore

Tuition discounting: New record high at private nonprofits—56.3% for first-time undergrads in 2024–25. That’s margin squeeze.

Enrollment mix: Spring 2025 enrollment is up 3.2% year over year but still ~2.4% below 2019 undergrad totals. Demand is fragile.

Closures and mergers: Trackers list 80+ closures/mergers since 2020 across public and nonprofit schools; new cases continue.

Student debt stress: After repayment restarted, delinquencies rose fast in 2025; outstanding student loan balances sit around $1.64 trillion (Q2 2025). That pain shapes public opinion and enrollment choices.

Ten Unusual or Unintended Consequences to Acknowledge

We may not go deep on these, but they’re worth flagging:

1. College-town economics: Fewer students → fewer renters → weaker demand for restaurants and shops.
2. Sports fallout: Program cuts mean team cuts; leagues reshuffle.
3. 529 plan shifts: Families redirect 529 education savings toward short programs, certificates, or wait longer to commit.
4. Accreditation pressure: Agencies face calls to approve competency-based and skills-first programs faster.
5. Testing arms race: “AI-proof” assessments grow; AI-vs-AI plagiarism checks become standard.
6. Insurance and liability: More rules around human-in-the-loop review in health, engineering, and finance.
7. Corporate mini-universities: Big employers build in-house academies and fund micro-credentials, bypassing colleges.
8. International student flows: If non-STEM demand falls, some schools lose a key revenue stream.
9. Civic life and mental health: Fewer campus “third places” could harm community ties for young adults.
10. Curriculum gatekeepers: A few AI platforms might quietly shape what people learn by default through templates and prompts.

Objections—and the Short Answers

“Enrollment is up. People still want degrees.”
True, enrollment rose in Spring 2025—but undergrad counts are still below 2019. The rebound doesn’t fix cost, debt, or job-market alignment.

“Skills-first is hype; degrees still rule.”
There’s hype and there’s change. Indeed shows fewer postings with degree requirements and more with none at all. HBS/Burning Glass document real “degree resets” in many middle-skill roles. Execution is uneven, but the trendline is clear.

“AI adds jobs too.”
Yes—new jobs appear while old tasks decline. But in the worst-case timeline, the timing gap hurts lecture-heavy programs before benefits spread.

“AI can’t do real work.”
“Computer use” features let models operate apps and browsers, not just chat. It’s early, but the direction is obvious.

The “AI Will Eat My Major” Heat Map (Plain English)

• High exposure right now: General business, communications, many social science tracks, non-lab IT/analytics, arts theory courses. These rely on essays, slide decks, briefs, and basic dashboards—the things AI now drafts well.
• Medium exposure: Computer Science (CS). Yes, AI writes code—but humans who ship, secure, integrate, and own systems still matter a lot.
• Lower exposure: Licensed clinical fields, hands-on engineering, field sciences, skilled trades. Physical risk, compliance, and real-world context still demand humans and physical assets (labs, clinics, equipment).

Playbook for Learners (Ages ~18–28)

1) Make AI your daily teammate.
Pick one top model and learn how to use it for writing, analysis, and simple code. Practice “prompt + check”: ask, get output, verify with sources, and revise.

2) Build a “Portfolio of Proof.”
Every 2–3 weeks, ship a finished artifact that employers value: a decision memo, market teardown, KPI (Key Performance Indicators) dashboard, prototype, case brief, or audit. Publish on GitHub, a site, or a portfolio page. (Your degree is a label. Your portfolio is proof.)

3) Show your process, not just the result.
Add a short note to each project: How did you use AI? How did you check it? Show your human-in-the-loop steps.

4) Seek apprenticeships and scoped projects.
Short projects with real stakeholders beat long lecture chains. Aim for 2–3 scoped projects each quarter.

5) Manage debt and risk.
Avoid large loans for majors without licensure, labs, or strong placement data. Compare Return on Investment (ROI): dollars in vs. job outcomes. If a program’s job data is vague, treat that as a warning.

6) Learn the tools of the job.
Pick tools that show up in postings (for example, a cloud platform, a data stack, a design suite). Build at least one real project with each.

7) Network the right way.
Lead with your work sample, not “Can I pick your brain?” Offer a Try-Me Task: “Give me a 2-hour problem; I’ll show you how I work with AI.”

Playbook for Degree-Holders (Early and Mid-Career)

• Refactor your résumé around deliverables: systems owned, KPIs moved, playbooks built, risks managed.
• Document “man + machine” workflows: show where you use AI, where you verify, how you handle edge cases.
• Run a 90-day skills sprint aimed at the roles you want: pick five job postings, map missing skills, and build two public artifacts that close the gap.
• Publish a living portfolio and add two scoped projects per quarter—especially in adjacent domains to widen your proof.

Playbook for Employers

• Move from pedigree to paid trial tasks (4–8 hours). Allow AI; score judgment and QA (Quality Assurance), not just output.
• Hire for interface skills: Can the person design a safe human-in-the-loop workflow? Can they catch hallucinations and own results?
• Build an internal academy: 6–12 week tracks tied to clear KPIs. Link completion to pay. (Plenty of firms say skills-first—far fewer do it. Close the gap with simple, repeatable task trials.).

Playbook for Colleges and Universities

1) Flip lectures into studios and clinics.
Grade production under supervision, not just tests and essays.

2) Write an AI-first syllabus.
Every course defines: allowed AI tools; audit trails; red-team checks (students test AI outputs for failure modes); and process grading (not just final results).

3) Double down on physical capital.
Where you have unique labs, clinics, or instruments, lean in. Price and market that moat.

4) Portfolio-based graduation.
Students leave with a public body of work: working prototypes, clinical write-ups, lab notebooks, field reports, and briefs.

5) Employer councils with teeth.
Quarterly job-task refresh → syllabus update → live projects in the Learning Management System (LMS) every term.

6) Be honest about risk.
If a program has no clear hiring path, post real numbers (placement, underemployment). Update or sunset it.

A Simple “Worst-Case Dashboard” to Watch

• Tuition discount rate. If it keeps rising, stress is growing.
• Closures/mergers per quarter (closure trackers). Rising pace = real trouble.
• Degree mentions in job posts vs. skills tests/portfolios (Indeed + HBS/Burning Glass). Fewer degree requirements = signal shift.
• Frontier AI capabilities (especially “computer use” and agentic coding benchmarks).
• Student loan stress: balances (~$1.64T in Q2 2025) and rising delinquencies. If borrowers feel squeezed, enrollment and public support weaken.

If You’re 18–28: A 90-Day Action Plan

Week 1–2

• Pick a domain (ops, analytics, design, growth, compliance, clinical support).
• Set up your AI stack (one general model + one coding helper + one research tool).
• Choose two tools from real job postings (for example, a cloud platform + a BI dashboard).

Week 3–10

• Ship four artifacts, each tied to a real job task:
• Decision memo with a simple model (assumptions + pros/cons).
• Market teardown with a one-page KPI dashboard.
• Prototype (no-code or code) that solves a tiny workflow.
• Compliance or safety brief showing risk thinking in your domain.
• For each: add a process note—how you used AI and how you checked it.

Week 11–12

• Send 10 outreach messages with a “Try-Me Task” (2–4 hours). Offer a clear deliverable and deadline.

Week 13

• Convert 2–3 of those trials into paid gigs, internships, or apprenticeships. Update your portfolio and repeat.

This outperforms many lecture-only paths on speed, cost, and proof.

Where Myford University Fits

We are building Myford University for the world this stress test imagines—even if reality lands softer.

• 8–12-Hour Accelerators (currently in development) that turn learning into deliverables you can ship:
• MBA Accelerator: decision memos, market tear-downs, KPI dashboards, cap table models.
• College Accelerator: analytical writing, structured problem-solving, research briefs.
• PhD-Level Thinking Toolkit: epistemology, research design, systems thinking, bias checks.
• Every accelerator ends with a portfolio artifact you can point to on Monday.

Join the Debate (Comment Below)

I ran this worst-case on purpose. Now I want your take.

Tell me where I’m wrong. Tell me-- if you were 18 today, would you enroll—or build a portfolio?

10 ideas for you to spark real discussion in your comment:

1. What would have to be true for your job to be 50% automated in three years?
2. If you’re 18–28: enroll, defer, apprentice, or start a business—what’s your move and why?
3. Which majors gain and which lose first—and what evidence backs your view?
4. What’s one lab or clinical experience AI can’t replace this decade?
5. If you hire: which work test beats a degree for your roles?
6. Which Key Performance Indicators (KPI) would you watch to call a college “at risk”?
7. If you run a college: what would you stop this year? What would you double down on?
8. What’s the fairest way to help students caught mid-degree if programs shrink or close?
9. Name one unintended consequence nobody’s talking about.
10. If you disagree with the thesis, what proof would change your mind?

Rules of engagement (short and clear):

• Be specific. Share data or examples when you can.
• Respect others. Attack ideas, not people.
• Keep it useful: one insight someone can use this week.

Myford University — learn fast, apply it faster.

Bottom Line

If your plan depends on lectures and essays, it’s living on borrowed time.
If your plan depends on labs, licensure, instruments, and production, you have a moat.
If your plan depends on proof over pedigree, you win either way.