Wharton Professor: Even If the AI Bubble Bursts, Jobs Won’t Go Back to the Way They Were

All through 2025, the debate around AI spilled out of the capital markets and into every leadership meeting.

Is it a bubble? Can it still make money? Should companies deploy it everywhere—or hold back?

In the middle of that noise, the podcast Factually! with Adam Conover released a conversation with Wharton professor Ethan Mollick—someone who didn’t just study AI from a distance, but spent an entire year using it in real classrooms.

Mollick has become one of the most visible “AI-in-practice” voices online. His research and teaching focus on how AI is reshaping work, education, and the paths people take to build companies and careers.

His position is straightforward.

Stop arguing in extremes—“AI replaces everything” versus “AI is useless.” The truth sits in the middle: AI won’t replace everything, and it’s far from meaningless.

That’s exactly what this piece is about—his central takeaway from the conversation:

Even if the AI bubble bursts, work won’t go back.

What’s changing isn’t simply whether a job disappears. It’s how work gets done.

1 | AI Isn’t Just a Tool. It’s a Collaborator.

For many people, “using AI” started with chat tools—ChatGPT, DeepSeek, Doubao, and others.

You type a question and get a few hundred words back. You ask it to draft an email, polish a resume, translate a report—and it helps, fast.

Some people feel amazed for three minutes. Others shrug and say, “So it’s just a smarter search box.”

Mollick sees it differently. In his view, this isn’t a tool upgrade—it’s the arrival of an additional teammate.

In his Wharton courses, he used AI continuously for a full year: delivering lessons, grading work, writing questions, answering student inquiries, and building teaching materials. His observation is blunt:

AI is available to everyone now. Whether you choose to use it is up to you. But if you don’t, it will still be helping someone else.

A tireless, on-demand brain is sitting in front of the entire workforce.

Mollick shared a simple example from his own writing process. While drafting academic work, he asked GPT to review parts of his reasoning—and it flagged a logical gap he hadn’t noticed. Students, meanwhile, used AI to create early drafts and case analyses, then discovered their own blind spots faster because the first pass already existed.

Not because the model “thinks” like a human.

But because it can take a messy, vague prompt and shape it into a workable starting plan.

Older AI felt like software sitting on your desktop.

Today’s AI feels more like a collaborator in your work chat—always available, always ready to produce a first version.

People are already using it to:

Write research summaries

Generate business strategy drafts

Build the first version of a new product

It’s not here to “take over” your work. It’s lowering the friction of getting work started—and raising the baseline speed of delivery.

And while some people are still watching from the sidelines, others are already shipping.

2 | AI Isn’t Replacing You—It’s Eating Your Role, Task by Task

The fear that keeps coming back is simple: “Will AI take my job?”

Mollick’s response is that the question misses the real mechanism of change.

AI doesn’t need to replace an entire job title overnight to transform the labor market.

Instead, it pulls individual tasks out of your role—one by one.

Take doctors. “Doctor” isn’t going to vanish in a flash. But many tasks inside the job are already being automated or accelerated: interpreting test results, drafting clinical summaries, flagging abnormal indicators, turning notes into structured documentation.

Mollick’s point is sharp:

AI doesn’t look at job titles. It looks at tasks.

Most modern roles are bundles of tasks. You do communication, write reports, gather information, refine language, summarize data, build slide decks, draft proposals—often all within the same week.

And now, many of those tasks can be handled independently by AI.

In the past, stable divisions of labor helped people build stable professional identities.

Now, roles are less stable. What matters more is your portfolio of capabilities—your ability to combine skills and adapt as the task mix shifts.

The later you recognize that, the harder it becomes to keep pace.

3 | You Think You “Use AI,” But Most People Use It Shallowly

You might feel like you’re already keeping up.

AI for emails. AI for summaries. AI to patch a section of a deck you don’t want to write. AI to clean up wording.

That’s real efficiency—no question.

But Mollick’s observation is that many people stop there. They treat AI like an emergency helper: fill in content, fix typos, polish sentences, make things sound nicer.

The people who get outsized value don’t just “use” AI. They pull it into the workflow as a real contributor.

He pointed to examples like these:

A World Bank experiment in Nigeria paired AI with teachers for after-class tutoring; after six weeks, students improved dramatically—roughly equivalent to gaining a full additional year of learning.

Research teams have found that programmers using AI coding tools increased code merge volume by about 39% without raising error rates.

At Wharton, students and AI systems each generated 200 startup ideas; when external judges selected the top 40, 35 came from AI.

That’s not “AI makes me faster.”

That’s “AI participates in every stage of the work.”

And there’s a counterintuitive pattern here.

The people who look like they’re “into AI” sometimes keep it confined to small chores.

Meanwhile, the people who seriously ask, “What can AI carry for me?” tend to restructure their process—and unlock far more leverage.

This is the dividing line:

Shallow use: AI helps you fill in content.

Deep use: AI produces the first draft.

Once the first draft exists, you stop starting from zero. You move straight into refinement, judgment, taste, and decision-making.

Knowing how to use AI is the entry ticket.

Using it deeply is where the gap opens.

4 | The Bubble Might Burst—But the Changes Won’t Reverse

Many people are waiting for a signal: “When does the AI bubble pop?”

Mollick’s answer:

A bubble can pop. AI won’t disappear.

His reasoning is practical. There are massive numbers of users already, multiple companies releasing models (including free ones), and major platforms like Google and Microsoft aren’t going to abandon AI because a stock narrative shifts.

Even if a company’s valuation collapses, the infrastructure doesn’t evaporate: data centers built, models released, workflows redesigned, user habits formed.

He compares it to the late-1990s internet bubble. The market crashed, but the fiber got laid—and the world still moved online afterward.

The deeper point is this:

Work habits have already shifted.

Once people experience a step-change in speed and output, it’s hard to willingly return to the old rhythm.

But what Mollick worries about most isn’t individual productivity—it’s the long-term damage to how people learn.

He calls it the risk of apprenticeship collapse.

For thousands of years, humans have passed skills through apprenticeship: juniors do the work, seniors correct them, and competence builds through guided repetition. White-collar work operates the same way—interns and new hires learn by doing under supervision.

But something changed last summer.

Mid-level managers became less willing to train interns, because AI could complete the work faster.

Interns, in turn, started handing the work to AI—because AI was genuinely better than they were.

The result is a dangerous loop: AI talking to AI, while young workers lose the chance to build real skill through practice.

Once that loop becomes normal, it’s extremely hard to return to the era of “hands-on teaching.”

Mollick repeats a warning throughout: the most risky stance isn’t fear of AI—it’s pretending it will go away.

The right move isn’t to wait for AI to vanish.

It’s to reduce the downsides and expand the upside—deliberately.

Because while you wait for the bubble to pop, someone else is already using these systems to build the future.

Will the bubble burst? It might.

Will work return to the way it was? It won’t.

What matters now is the choice in front of you:

In a world that has already changed, where do you want to stand?