Conviction's Sarah Guo on the SaaSpocalypse, AI reskilling as a venture opportunity, and the coming ChatGPT moment for robotics

Summary

Sarah Guo, managing partner at Conviction, sees the SaaSpocalypse not as a cyclical dip but as a structural rupture. Enterprise software companies that built durable businesses at $100M+ ARR before ChatGPT and have not dramatically overhauled their strategy over the past three to four years face a genuinely existential problem. She is not predicting consolidation into a handful of labs, though. Portfolio companies like Layton Health, Harvey, and Sierra occupy domains that are clearly not in the pathway of what frontier labs are building or want to build.

Reviving a Legacy Software Company Requires Three Simultaneous Moves

Guo frames the AI transition for mature software companies around three concurrent requirements. First, leadership with moral authority, typically a returning founder, must install genuine urgency. Second, AI adoption cannot be siloed into a single crack team. It needs to reach every function, finance, sales, product, and strategy, through hackathons, internal experimentation norms, or company-wide mandates. Third, companies must build a compounding AI-driven motion large enough to enable innovator's dilemma decisions on pricing and investment allocation. She cites Simon at Notion as an example of founder-led execution and flags acquisition as a fallback when internal talent is lacking.

Workforce Bottlenecks Are as Real as Compute Bottlenecks

Guo argues that the hard constraints on U.S. AI capacity expansion are physical and human, not algorithmic. Brad from OpenAI has flagged that the bottlenecks to increased AI infrastructure are centered on labor, specifically electricians, data center technicians, and power generation workers, all areas where Washington has direct policy levers. Even with aggressive reskilling investment across startups, universities, and government programs, she estimates the country needs roughly ten times as many electricians as it can realistically train in the near term. The only viable path to closing that gap at speed is automation.

Guo discloses she is a long-standing investor in Uplimit, a company focused on AI-age reskilling that currently serves enterprise clients through digital programs rather than trades. She frames reskilling as a venture-scale opportunity but is explicit that it is insufficient on its own given the pace of capacity demand.

Robotics Is Two Years From Its ChatGPT Moment

Guo expects a consumer inflection point in physical AI within the next two years. Waymo is the first mass-market exposure to autonomous robotics, and consumer acceptance is tracking the familiar S-curve: skepticism on first contact, normalisation on second, and impatience for performance improvements by the third ride. She sits on the board of Sunday Robotics and expects robots entering the home to be the moment that mainstream consumers accept general-purpose robotic capability, alongside parallel penetration in manufacturing and industrial environments. She draws a direct parallel to ChatGPT's current 900 million users as the scale benchmark robotics is trending toward.

The Next Big Consumer AI Product Has Not Appeared Yet

Guo endorses Ben Thompson's characterisation of OpenAI as an accidental consumer company, a research lab that stumbled into a mass product. She considers the OpenClaw agent, which combines memory, persistence, and the ability to act on a user's behalf, a leading signal of what comes next. That category still has not penetrated mainstream consumers despite strong enthusiasm in tech circles, which she reads as evidence that the consumer AI application layer remains wide open for a new breakout product from an unexpected source.