Khosla’s Ethan Choi On AI, Founder-First Investing And The Fate Of Entry-Level Jobs

As a partner at Khosla Ventures, Ethan Choi isn’t shy about speaking his mind.

The investor is vocal about his belief that AI is a massive threat to entry-level jobs, or what he views as a shifting social contract in the modern workforce.

He would know about AI’s impact. Choi has led investments in several AI-first companies.

Known for a founder-first conviction, he’s also backed enterprise software and fintech infrastructure companies.

Prior to joining Khosla in 2024, Choi was a partner at Accel, where he led and managed high-profile growth investments in companies such as 1Password, Klaviyo, Pismo (which was acquired by Visa), Nuvemshop and commercetools. Before Accel, he worked at Spectrum Equity, backing companies such as Lynda.com (acquired by LinkedIn), Headspace, Lucid Software and PicMonkey (acquired by Shutterstock).

I recently spoke with Choi to hear more about why he thinks entry-level jobs could be disappearing, why he’s flipped his investing philosophy, and how he’s gone from growth-stage investing to being stage-agnostic.

This interview was edited for brevity and clarity.

Ethan Choi, partner at Khosla Ventures. (Courtesy photo)

Crunchbase News: You’ve had a prolific run the past couple of years, leading deals in Ramp, ClickHouse, Vercel, Glean, Bridge and others. How are you managing that volume?

Choi: It has been an intense stretch. About seven deals happened just last year, which was an insane year by any standard. This year has been a bit calmer as I focus on settling in with the companies I’ve invested in.

You’re currently researching the disappearance of entry-level jobs. As a parent, that sounds a bit scary to me. What are you seeing?

It’s a massive conundrum. I’m seeing it in my own workflow. I use the models to get up to speed on technical capabilities — asking about Clickhouse’s indexing methodology versus Snowflake’s in voice mode while I’m driving. It’s pulling from research papers and documentation faster than any human could. I describe it as feeling like I have an “Ironman suit” on.

The problem is that if I can do the work of a junior associate myself, almost instantly, those roles vanish. We’re facing a world where the base-level work we used to rely on young folks for is now table stakes.

If the “on-the-job” training era is over, where does that leave students and universities?

The burden of the first three years of “learning how to be a professional” has to shift to the universities. I look at the traditional U.S. model of general education requirements and think, “Why are we doing this?” We did that in high school. Universities should be places where you use AI to actually build things and apply knowledge to the real world.

If you’re a computer science major today, you need to graduate looking and delivering like a third- or fourth-year engineer. The bar has been raised for everyone. While schools like Vanderbilt and Arizona State University are leaning into an “AI-first” curriculum, many elite institutions are still silent, trying to figure out how to adapt.

Khosla is known for being contrarian. How does that translate to the growth stage in such a competitive market?

I’ve actually evolved to be stage-agnostic. While people still put me in the “growth” bucket, I’m doing much more seed and Series A. In this era, the metrics a company has today don’t guarantee where they’ll be in two years because the rate of change is so high.

I’ve flipped my philosophy: it used to be 80% metrics and 20% founders. Now, it’s 90% founders. The only constant is how special the founding team is and how quickly they can adapt. If code is being created 10x faster, a company might face 50 years of change in a single decade. You have to back the people who can handle that stress.

You’ve predicted “mass carnage” for some software companies. Who survives the transition to an AI-native world?

We are moving from trading on revenue multiples to trading on free cash flow and PE multiples. That’s a painful transition. The market now needs to believe that a company is AI-native — that its revenue is moving toward inference- and usage-based models rather than just old-school seat licenses.

I expect carnage for lightweight, horizontal applications and mid-market companies that can’t attract applied AI talent. I have a ton of respect for founders like those at Intercom or Airtable who are “burning the boats” to reinvent their entire businesses. It’s incredibly difficult, but in this market you either reinvent or you get replaced by someone building natively from day one.

With your background in fintech infrastructure, where do you see the next “unconventional” opportunity in financial services that most growth investors are currently overlooking?

It’s now become somewhat consensus, but I still believe it’s fairly unconventional that systems of record can be ripped out, whether it be in financial services or in other categories.

For example, we recently invested in DualEntry, which is seeking to replace NetSuite and the core accounting system, which is the last system of record I would have thought might be under threat. We’re seeing that with AI, startups can build migration paths that didn’t exist before, and also the depth and breadth of incumbent platforms in a fraction of the time.

Vinod Khosla often talks about “challenging the conventional wisdom” of founders. Can you share an example of a time you had to steer a growth-stage founder away from a “safe” path toward a much larger, albeit riskier, vision?

In general, there are many times when a founder is thinking through a very risky but potentially game-changing product addition or acquisition. While I view part of our job as investors and board members as helping identify and manage potential risks, the most important thing we can do is give them the courage to take risks that are transformational to the business and the category they are in.

You’ve noted that talent density is the most important variable for success. In a market where AI is automating routine work, how has your criteria for what defines an “elite” executive hire changed?

Perhaps somewhat ironically, one of the main differences in criteria is whether this exec has “IC’d” (individual contributor’d) themselves and can do most of the work required out of the gate with their own two hands plus AI.