Sarah Guo is the founder of Conviction, an AI-focused venture capital firm backing companies including Baseten, Harvey, Mistral, Cognition, OpenEvidence, and Sierra. Her defining investment thesis is that advances in artificial intelligence become durable businesses through exceptional products, intimate knowledge of customers, and relentless execution.
Guo grew up around her parents’ technology company, built its first website as a teenager, and studied economics, literature, and statistics at the University of Pennsylvania. After working at Goldman Sachs, she joined Greylock in 2013 and became a general partner in 2018, investing across enterprise software, infrastructure, security, and machine learning.
She founded Conviction in 2022 with an initial fund of approximately $100 million, shortly before ChatGPT’s arrival accelerated the AI startup market. Her earlier backing of Baseten continued into Conviction, connecting her experience in infrastructure investing with the firm’s broader portfolio across foundation models, developer tools, legal software, clinical information, and customer service. Former Sequoia investor Mike Vernal joined as a general partner in 2025.
Conviction also runs Embed, a program supporting early-stage AI founders. Guo co-hosts No Priors with Elad Gil, interviewing researchers and founders about AI’s technical and commercial development.
What she looks for in AI companies
- Execution is the moat. Guo sees customer context, model orchestration, interface design, reliability, and trust as defensible advantages even when competitors can access the same underlying models. Cursor exemplifies that approach: it integrates retrieval, multiple models, familiar developer workflows, and transparent code changes into a coherent product.
- Cursor for X. Coding assistants succeeded partly because engineers understood the work they were redesigning. Guo expects comparable opportunities in law, medicine, customer support, and other professions when founders build around actual workflows and supply relevant context automatically.
- Human-guided copilots. She defines agents by planning, goal retention, task ownership, and the ability to try alternative approaches, while recognizing that users often prefer augmentation when latency, errors, and consequential decisions make full autonomy impractical.
- Product taste and operating pace. Guo’s writing on product taste extends craft beyond interfaces to hiring, sales, and customer trust. Her argument for operating pace emphasizes prioritization, rapid feedback, and decisions made despite incomplete information.
- Opportunities beyond internet text. Voice and video open business workflows that text alone serves poorly; robotics, biology, materials science, and physical simulation demand specialized data and domain expertise unavailable from the public web.
