No Priors Ep. 72 | With Sarah Guo and Elad Gil

Detailed Summary

Introduction (0:00 - 0:16)

• The hosts, Sarah Guo and Elad Gil, introduce a host-only episode of "No Priors."
• They begin by discussing a recent Goldman Sachs report titled "Calling the Top on AI."

Old School Operators' AI Misunderstandings (0:16 - 5:10)

• Sarah criticizes the Goldman Sachs report, particularly arguments from Darren Asoglu (MIT professor) and Jim Cavello (GS head of global equity research).
• Asoglu claims AI will impact less than 5% of tasks and that trillions in capex for model training are a waste, arguing AI can't solve complex problems.
• Cavello suggests AI is too expensive initially and that efficiency gains will be competed away, unlike the internet which disrupted expensive solutions from the start.
• Elad and Sarah argue these views stem from a misunderstanding of modern AI (Transformer-based models) and its scale-dependent improvements, contrasting it with "old school ML."
• They emphasize that AI capabilities are rapidly improving and becoming cheaper, making long-term negative predictions ludicrous.

Tech History Repeating Itself with Slow AI Adoption (5:10 - 6:09)

• Elad highlights that enterprise AI adoption is still in its early stages, primarily focusing on vendor tools and internal applications, with customer-centric products just beginning to emerge.
• He draws a parallel to the internet era, where initial skepticism from established companies (like IBM's CEO calling internet companies "fireflies") eventually gave way to massive market shifts.
• Sarah notes that the assumption that AI advantages will be standard across all companies is flawed, as some will become "Amazon" while others become "Borders" in the new AI landscape.

New AI Markets (6:09 - 8:48)

• Elad identifies three drivers for new AI markets: new generalizable capabilities across various fields (biology, robotics, language, image), API accessibility allowing anyone to use AI, and a widespread corporate mandate for AI adoption.
• He points out that while software spend is around half a trillion dollars annually, services industries have 5 trillion in headcount costs that AI could transform (e.g., legal, accounting, sales).
• This environment makes incubation and AI-driven buyouts viable strategies for the first time.

AI-Backed Buyouts (8:48 - 13:03)

• Sarah explains that AI-backed buyouts shortcut change management and technology adoption in industries that are ripe for automation but slow to adapt.
• Elad elaborates that these buyouts involve taking over an asset to entirely rework its functions with AI, citing Clara's customer support team reducing headcount by 700 while improving service.
• He distinguishes this from traditional "tech-enabled buyouts," which often involved arbitraging valuations with a thin software veneer; AI buyouts offer dramatically higher technological leverage.
• The success of these ventures requires a blend of software founder expertise and private equity operational skills, rather than just one or the other.

AI Incubation (13:03 - 17:18)

• Elad discusses that while incubations are generally risky, they make sense when there's proprietary visibility into an opportunity or customer set, and the required expertise is unlikely to form organically.
• He notes that successful incubators often have deep vertical expertise, relationships, and captive customer bases.
• The current AI landscape presents unique opportunities for incubation due to market crowding in some areas and wide-open spaces in others, coupled with strong enterprise interest.
• Elad mentions working on several incubations, including one with Ankur Goyal (Brain Trust), driven by clear market opportunities and customer pull.
• Sarah adds that incubations are particularly effective now because technologists understanding AI often lack domain expertise, and vice versa, creating a mismatch that incubators can bridge.

Exciting Incubating Applications (17:18 - 18:26)

• Elad shares that he is actively looking into healthcare-related AI incubations, leveraging large enterprise assets for instant customers.
• He is also exploring opportunities in the services world and for new types of large-scale models for specific applications.
• Sarah expresses excitement about a recent healthcare-related project.

AI and the Public Markets (18:26 - 22:20)

• The hosts discuss how AI impacts public market investments and portfolio construction.
• Sarah suggests shorting mid-to-late-stage private or small public companies that lack the will to adapt to AI.
• Elad outlines four segments for public market consideration: the "new Magnificent Seven" (companies with multi-hundred-billion or trillion-dollar potential like SpaceX, Stripe), pre-existing big tech companies (e.g., Apple with an AI-driven iPhone replacement cycle), "AI durable" companies where AI doesn't significantly alter their competitive landscape, and new AI index plays beyond Nvidia.

Staffing AI Companies (22:20 - 25:14)

• Elad notes that current founders are highly motivated and ambitious, aiming for high revenue per employee and small teams.
• Sarah believes this trend towards smaller teams is an extension of the startup ethos of solving problems with limited resources, amplified by technical capabilities.
• Elad suggests that while AI offers efficiency gains, significant headcount reduction from coding co-pilots is still a few years away, waiting for AI coding agents to work at scale.
• He questions when the "GPT generation equivalent" will hit enough of the economy or tech companies to automate significant portions of a startup's functions.
• Sarah agrees that current models aren't coherent enough to eliminate headcount in the first 10 people of a startup, but functions like SDRs and support are expected to see headcount shrink first, likely in two to three years.

Competition and Shrinking Headcount (25:14 - 29:06)

• Sarah observes that the pace of competition and change in the tech environment has accelerated due to underlying technology advancements.
• This aggressive race emphasizes velocity, creating a trade-off between hiring experienced leaders (who may have a slower pace) and maintaining product velocity.
• She suggests that AI-driven companies with market pull can prioritize velocity over immediately bending to experienced leaders, or be more selective in finding leaders who match their pace.
• Elad adds that AI companies, particularly in creative fields, are often more founder-driven in their product and model design, reflecting the founder's aesthetic taste and choice of data, making the "model design" a crucial part of the product experience.

Notable Quotes

"AI will be unable to solve the complex problems it's it's not built to do that." — Darren Asoglu (quoted by Sarah Guo)

"This is the early days and if you look at Enterprise adoption of AI most large Enterprises are very early on." — Elad Gil

"You are shortcutting the change management process for industries that can be like dramatically automated with these new capabilities." — Sarah Guo

"Incubations usually are a terrible idea." — Elad Gil

"The set of people who understand the technology are not commonly the set of people who understand the domain." — Sarah Guo

"I think there is a mid to late stage private company small public company that does not have the speed and institutional and founder leadership will to go change the business when they see the writing on the wall." — Sarah Guo

"Models are such a big piece of the product experience and that the uh thus the like model design... is quite important and often comes from the founder." — Elad Gil

"The pace of change of the underlying technology has accelerated such that um you know you have this like increasingly aggressive race." — Sarah Guo

"This is not a new story during the internet era there were people who said that the internet was meaningless." — Elad Gil

"The internet companies were fireflies before the storm." — CEO of IBM (quoted by Elad Gil)