Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)

The classic PM role will not survive as-is. The winning skill is closer to a CEO’s judgment about what to build and why, in a market where AI will upend careers and compress roadmaps to weeks.

You’ve got spicy takes lined up, from not talking to customers to how feedback should land. But first, you’ve helped more teams hit world-class than almost anyone I know.

My guest is Keith Rabois, operator and investor behind companies like Stripe, Airbnb, YouTube, and DoorDash. Before we dive in, you’re dialing in on an iPad and you said you have not used a computer in years—what’s the story?

I switched to an iPad in 2010 after seeing Jack Dorsey run on one and never looked back. I do everything from my phone, watch, or iPad.

Engineers are now coding from phones with AI as a partner, which makes your setup feel prophetic.

It cuts distraction and adds flexibility; I carry it everywhere. Unless you need heavy-duty engineering, the lighter tool wins.

Let’s talk about building elite teams in a weird hiring market where jobs are scarce yet top talent is harder than ever to attract. What do the best get right?

The team you build is the company you build, full stop. PayPal worked because Peter and Max concentrated exceptional people; everything else got easier.

Early on I was average at hiring strangers, so I recruited underused stars internally and proved leverage. Over time I learned to do ruthless references and to ask the right questions, because framing can flip a decision from no to yes.

What questions surface real signal?

With candidates I ask, if you ran your last company, what would you have done differently—and why could you not persuade the CEO? In references I ask what conditions maximize their success, and what root cause would explain failure.

How do you tighten the hiring feedback loop?

Ask the whole panel at day thirty if they’d rehire; that early read is as predictive as waiting a year.

Walk us through your barrels and ammunition model.

Most companies add headcount and stall because only a few people can carry an initiative from zero to impact. Those are barrels; everyone else is ammo, and the barrel-to-ammo ratio sets how many big bets you can run.

Ammo still matters, right?

Absolutely, but you size it to the problem; structure follows the right ratio, not headcount vanity.

How do you spot a true barrel?

Point at a hill and they get you over it—rallying people, securing resources, measuring progress, and surfacing issues only when needed. They’re fire-and-forget on outcomes.

Give us a concrete example.

At Square I wanted cold 9 p.m. smoothies for engineers and could not get it done; an intern solved it on day two and never missed again. That ownership earned him bigger and bigger missions.

When everyone is chasing the same stars, how do you convince a barrel to join you?

Sell mission, then show how their superpower unblocks your biggest constraint so they’re betting on themselves; that is how I left Google for Square. Long term, win on undiscovered talent others misjudge.

Big firms run homogenous filters, so understand why they will miss someone and pounce; less credentialed or younger candidates often have fewer data points and more upside.

People say you raise the bar harder the better the team performs—true?

The CEO’s job is sustained pressure; be supportive when the team is struggling and intensely critical when it is winning to fight complacency. Your best people hate coasting and morale drops when the pace slows.

With AI reshaping work, what is your career advice?

Stay relentlessly curious; in top orgs the CMO often consumes the most tokens because they dive in and ship analytics or campaigns themselves instead of routing through layers.

What happens to the product triad of PM, design, and engineering?

Static, year-long roadmaps make no sense in a world where core capabilities change weekly. The winning org notices what is newly possible and ships next week, which weakens the case for a traditional PM intermediary.

But someone still has to decide what to build and align the team.

Across roles, the craft becomes CEO-like judgment on what to build and why; think chef as editor and commercial strategist more than line cook. Great engineers with business instincts will be especially valuable as AI becomes their second team.

A director of engineering I know manages twenty and ships as much code as he did solo by orchestrating AI like a parallel squad.

How do you see design evolving?

Design and code are fusing, and execution speed is the norm—Shopify requires live, working demos, not slideware. The real edge is storytelling that slices through noise.

Distribution feels like the other end of the barbell now.

Cutting through real-life clutter matters more than lab tests; interviews are poor proxies for attention in the wild. Even simulated user models are risky if trained on the wrong data.

Do you buy that AI content will surpass human output?

We’ll see two tracks: authenticated human work with provenance, and algorithmically ranked best content regardless of source. LLMs shine on short form due to token economics, and the first lines often decide the outcome.

Your legal past—helpful or harmful in startups?

Useful for regulated bets and risk tradeoffs, but law trains you to spot issues and bill hours, which clashes with creating and shipping.

One of your enduring hot takes: avoid talking to customers unless you sell into the enterprise.

Consumer and SMB interviews mislead because people cannot explain subconscious choices; you anchor on bad data. Enterprise is different when you can reach the true buyer and map utility and politics.

So lean on conviction and learn from the market, not panels.

Study human nature, ship like you are releasing a movie, and make sure tickets sell; then fix the trailer and distribution before you rewrite the script.

Any examples of where instincts beat interviews?

DoorDash worked by solving density and unit economics, not by polling who wanted a button; Airbnb had early search signals on Craigslist that hinted at real demand. As an investor, I like ugly babies—ideas peers laugh at often carry the upside.

How do you evaluate this AI startup wave?

Ask whether foundation labs will crowd you out over a decade, and where compounding advantages will accrue; at seed I want founders who can articulate that path, even if they have not proved it yet.

What convinces you to invest early?

I am founder-driven: if there is a real chance this person can reshape an industry, I am in.

What do your best performers do differently day to day?

They move with startling tempo—identify a problem and have a measured fix by the next board meeting—and they compound that speed. I preempted Ramp’s A because they neared card launch in about three months, deepened talent density fast, and promoted from within instead of hiring lots of senior outsiders.

So that third pillar is growing leaders internally instead of importing star VPs, right? You keep stressing talent density, developing people into the job, and how that fuels speed.

We made that a strategy: default to promote from within. If a role is about preserving value, experience helps; if it is about creating value, you can often cultivate it.

I have seen chief of staff roles become a talent engine. One standout company turned chiefs of staff into a top CMO and a likely head of product by giving them real reps over a couple years.

On speed, Ramp lives it. They even track days since launch in public and treat velocity like a product.

Every board meeting there opens with the day count. It keeps urgency front and center.

Last hot take: should criticism happen in private or out in the open?

Do it in front of the team when you can. It aligns everyone that a problem is real and being handled, calms speculation, and invites others to jump in and help.

How do you avoid making that feel harsh or scary?

Use judgment like great coaches do, with some one‑on‑one when needed, but I do not optimize for psychological safety. I optimize for winning—think Jordan Rules or The Last Dance.

So you see psychological safety and sustained success as mostly at odds?

For the most part, yes.

Failure corner: share a miss that changed how you operate.

Our Google acquihire never matched our ambition despite a decent sale; in venture, even a great hit rate means many misses. Do not obsess over failure ratios—learn what mattered, keep people taking bold shots, and stay focused on winning.

Lightning round.

Let’s do it.

Go‑to book recommendations?

The Upside of Stress by Kelly McGonigal. Used well, more stress can make you healthier, happier, and more effective, and the evidence is convincing.

Recent movie or show you loved?

Nuremberg. It is not flashy, but it is rich with timely lessons, and I learned several new things; you can find it on Netflix or iTunes.

A product you are into right now?

Eight Sleep. I am an investor, but I truly believe in protecting eight hours, and this product helps people do it.

Do you have a life motto?

No days off, inspired by Belichick. I work and train every day.

So literally seven days a week?

Yes. I have missed seven workouts in seven years across travel and illness, and it is really about having no excuses.

You hit Barry’s this morning and have another later plus a lift. Final one: among the PayPal Mafia, who wildly exceeded your expectations?

No one surprised me that way; I already backed the folks I believed would build enduring companies.

Where can people find you, and how can listeners be helpful?

edge.com and on Twitter, where I post a lot; start with my pinned tweet.

Keith, thank you—this will help so many builders. If you enjoyed this, subscribe on Apple Podcasts, Spotify, or wherever you listen, and find past episodes at lennyspodcast.com. See you next time.