Teaching Entrepreneurship Will By no means Be The Similar

This submit beforehand appeared in Poets and Quants.

15 years in the past, my Lean LaunchPad class modified how entrepreneurship is taught. The class is now taught in tons of of universities worldwide and helped launch 1000’s of startups. But this previous summer season, I acquired fascinated with whether or not AI killed our Lean LaunchPad class, and with it the Lean Startup and Customer Development.

I spotted that if we needed college students to discover ways to construct companies fairly than AI slop, we needed to rethink the category.

Here’s what occurred, how we identified the adjustments wanted, and what we’re planning on doing going ahead.

---

For the final 15 years the cadence of the Lean LaunchPad class has been the identical. Students arrived with hypotheses of their product concept and goal clients. Each week scholar groups acquired out of the constructing and talked to 10-15 stakeholders. The following week, they offered “Here’s what we thought, right here’s what we did, right here’s what we realized, and right here’s what we’re going to do subsequent week.” Over the 10-week quarter, groups would discuss to 100+ stakeholders and use this suggestions to validate, modify or invalidate their hypotheses about their corporate affairs and refine and iterate their Minimal Viable Product in seek for product/market match.

The expertise was meant to reflect the real-world journey of a startup founder, and whatever the technical fad of the second (social media, cell apps, vitality, life science, protection, et al), this pedagogy has labored like clockwork for 15 years. Coming into class in Spring 2026, I had no concept that this 12 months can be its final, and the way educating entrepreneurship would by no means be the identical.

The first week of sophistication is all the time thrilling
Several months earlier than the Spring 2026 class was to begin, we interviewed the groups to listen to what issues they needed to work on and choose which might be a part of the category. In the primary official class, it was all the time fascinating to see how a lot they’d dug into the issue earlier than the primary day of the category.

In the primary class, groups usually offered a PowerPoint or wireframe of their product idea. This 12 months, nonetheless, when the primary workforce offered, there have been no PowerPoint or wireframe prototypes. Instead they demoed a full product (for Pediatric Sleep Apnea, Freight Forwarding, 3D-printed cooling for GPUs, music attribution, …)  Wow.

We had been impressed–this was the primary time we had ever seen a workforce develop one thing this quick and feature-rich on day one. As I used to be nonetheless processing what an ideal job this workforce had executed, the following workforce acquired up and in addition demoed a completed product. This time I used to be stunned. Two in a row!? By the time the three^rd, 4^th, 5^th, 6^th, 7^th and eight^th groups offered their full merchandise, all of us on the educating workforce had been surprised.

We saved one another making an attempt to substantiate, “Did you see what I noticed?”

To be trustworthy, at first we instructors had been giddy. All the groups had used AI to construct apps, digital twins or medical endpoints that in earlier years we might have hoped to see on the finish of week 10. We left class considering these groups had been on an ideal trajectory and thought for certain this begin would result in wonderful outcomes for all of them.

We had been incorrect, incorrect, incorrect.

As a educating workforce, we had been so enamored with this phantom progress that we didn’t cease the presses and refocus the scholars quick sufficient. We didn’t notice that AI had set our class on fireplace and would burn it to the bottom.

What AI modified contained in the classroom
In previous years, college students would get of the constructing and spend time making an attempt to deeply perceive clients’ issues. They used the corporate affairs mannequin canvas to seize what they realized as they examined all their hypotheses (go-to-market technique, pricing, product/market match, turnover, prices, and many others.) — all of the important components wanted to show an concept into corporate affairs.

AI made that course of fail.

Students used AI as an alternative of consumers for insights and validation. And as a result of this info got here from AI, they assumed it was appropriate.

AI made it simpler and quicker for college students to translate concepts into merchandise—however they’d no concept if this AI product met a buyer want or solved a buyer’s downside.

As the weeks glided by, the groups that had appeared so promising had been studying much less. Minimal Viable Products (MVPs) turned gross sales pitches as an alternative of experiments; groups collected compliments as an alternative of disconfirming proof; interviewees reacted to the product fairly than explaining their issues/wants.

Meanwhile, groups had been shocked to find that many potential clients had been already utilizing the identical AI instruments to create their very own alternate options simply as quick as they might. (This little bit of discovery was a sign to the workforce of what the ground was for options their startup may promote.)

In the tip, many college students couldn’t let go of the preliminary concepts that AI had helped them construct. Pivots turn out to be dearer psychologically, and people preliminary concepts turned frozen no matter proof they heard from clients. This was ironic, given pivoting the product was now technically low-cost. The educating workforce needed to intervene to pry these Initial Untested Products (what we had began calling MVPs) out of scholars’ palms.

The After-Action Review (AAR)
Just a few days after the category, whereas our recollections had been nonetheless recent, we gathered the educating workforce and Stanford school to share notes about what occurred, why it occurred, and learn how to enhance.

As we went across the room describing what we had seen and what we thought it meant, a couple of issues turned clear.

The impression on studying far outweighed the advantages AI supplied. To ensure there have been optimistic components of utilizing AI within the class. Students had constructed these wonderful merchandise utilizing Claude Skills and Gemini Gems. There had been tons of untapped alternatives to construct digital twins or take a look at 10s or 100s of apps concurrently. The impression on buyer discovery was equally spectacular. Assisted by AI, groups had been in a position to floor the proper inquiries to ask of the proper folks to get higher solutions to check their hypotheses quicker. Teams used ChatGPT for market analysis and Replit to construct web sites, Granola and Twinmind for notetaking; created artificial customers with Listen Labs and Viewpoints AI to check towards actual buyer knowledge; summarized their analysis in Google NotebookLM or Notion, then used Perplexity to create their weekly displays.

The MVP Is Dead
What was instantly apparent was AI’s impression on the Minimal Viable Product (an MVP). In the previous, an MVP was painfully developed, reflecting the week-to-week cumulative information gathered by speaking to stakeholders. An MVP additionally was proof of a workforce’s technical competence.

It struck us that having a product on day one meant an MVP was not proof of something: not buyer discovery, important considering, speculation testing, product/market match, buyer validation and even dedication.

Creating merchandise quickly at nearly no value had allowed groups to make dangerous concepts go quicker.

AI had created proof theater. These Initial Untested Products felt like proof however had been constructed with minimal or no contact with clients. They appeared like progress however dramatically raised affirmation bias and delayed pivots.

Student studying was unbalanced. A finished-looking product felt like success. Students confused a elegant deliverable with the necessity to deeply perceive the wants of all of the stakeholders, in addition to the seek for Customer Validation. Team understanding was much less nuanced; there was much less depth uniformly throughout the groups about the issue they had been fixing and the way nicely they understood buyer wants.

It wasn’t that AI was hallucinating – the groups had been. If they pivoted in any respect, they pivoted late as they assumed {that a} polished product meant product/market match. (Pre-AI groups pivoted 3-4 occasions.) One workforce did go “IUP loopy” and created new IUPs weekly whereas by no means letting their realized proof mount up.

All this added as much as studying debt. These Initial Untested Products (IUPs) let groups skip the battle which prior to now had led to buyer perception and understanding. The code labored, the deck was polished, and the evaluation was coherent, however by utilizing AI to summarize their interviews, groups missed the client insights. As a end result, they might not defend the assumptions or clarify the sting circumstances.

As the educating workforce mentioned what we had seen and what we thought it meant, a couple of issues turned clear.

1. The MVP as an artifact of studying a couple of worth proposition was useless.
2. The bottleneck in startups has moved from the time and price of constructing a product to judgment about what to construct and who to construct it for. This means founders nonetheless have to know which downside issues, who can pay, learn how to distribute, and learn how to transfer quicker than the opposite groups who can even construct one thing in a weekend.
3. When everybody can construct rapidly, what you select to construct and for whom turns into the entire recreation.
4. This means the aggressive panorama is way more necessary. Previously a workforce may spend a semester largely ignoring opponents as a result of the time and price of constructing a product turned a moat. That moat not exists. Teams now want a deep understanding of the present aggressive panorama and the speedy aggressive developments.
5. AI has made clients extra refined– now they’ll use AI to construct options as quick as startups can. This means the invention course of now additionally wants to seek out moats and paths to scale.
6. There are low boundaries to cloning. That similar ease of creation means startups have to discover ways to construct defensible moats — and to deal with a moat as a discovery downside, not a slide within the fundraising deck. IP was once defensible. Now what’s defensible IP?

Leaving the After-Action Review, we thought we understood the issue. The Minimum Viable Product was not a helpful artifact for studying and discovery in regards to the worth proposition and product/market match. I felt assured that we may make some easy fixes to the syllabus to take care of this.

[Insert laughter here.]

Much like our college students, we had simply confused the signs (the MVP is useless) with a lot, a lot bigger actual issues. After plenty of iterations we found the basis causes and learn how to preserve the category recent for the AI age.

Part 2 describes what we realized.

Filed beneath: Lean LaunchPad, Teaching |