The future of artificial intelligence applications will depend on going beyond automating existing tasks to fundamentally enhancing business models and accelerating human discovery. This was a central theme articulated by Andreessen Horowitz partners Oliver Hsu, Brian Kim, and David Harbor in their recent Big Ideas for 2026 discussion, where they outlined three key vectors that will define the next stage of AI: autonomous science, consumer product connectivity, and durable economic defensibility. Their analysis suggests that the true value of AI is unlocked when it moves beyond basic productivity gains to drive net new revenue and smarter outcomes.
Oliver Hsu, a partner focused on American Dynamism, introduced the concept of autonomous laboratories and argued that advances in AI reasoning and robotic manipulation are propelling scientific discovery toward closed-loop systems. Laboratory automation itself is not new. Pre-programmed robots have been handling repetitive tasks for a long time. Hsu explained that this change is a combination of advanced AI reasoning capabilities and physical automation that enable complex experimental design and iteration. “As model capabilities advance across modalities and robot operational capabilities continue to improve, teams will accelerate their pursuit of autonomous scientific discovery,” Hsu said, painting a picture of an AI scientist who can design, run, and learn experiments without continuous human input.
In the short term, this means collaboration. This means human scientists will work directly with AI systems that handle experimental workflows. Critical to this transition is interpretability. Because AI systems act as non-deterministic computers, researchers need to understand why the system plans experiments in a certain way to ensure scientific rigor and reproducibility. A focus on verifiable process records is essential for trust and ultimately full autonomy.
I believe these fields, especially life sciences and chemistry, are ripe for early adoption because they feature mature demand-side markets that are willing to pay for successful research results. The increased speed and capabilities provided by AI-driven labs, combined with public and private initiatives like the Genesis mission, are laying the foundation for this self-driving science to become a reality.
AI Applications Partner Brian Kim shifted the focus to the consumer environment, arguing that major AI products will pivot from mere productivity to true connectivity and identity. While the first wave of large-scale language models focused on helping users “do their jobs better,” the next generation will focus on making people feel “seen” and building stronger relationships. This includes leveraging AI to deeply understand users, capturing their digital footprint, communications, and history to foster meaningful relationships.
Kim argued that startups are uniquely positioned to compete with established platforms in this change. Incumbents have existing network effects and data moats, but AI offers “completely new user interactions” that may not fit natively into established platforms. This disruption provides agile companies with an outlet for creativity and an opportunity to build new atomic units of interaction that foster connection. Kim expressed excitement about the next wave of products, saying, “AI doesn't just help us do our jobs; it helps us see ourselves more clearly and build stronger relationships.” This personalization, driven by AI and deep shared self-knowledge, is the key to unlocking the next huge consumer market.
Finally, David Haber, also a general partner in AI applications, offered the toughest economic insight. The most durable AI companies are those where AI enhances the business model itself, driving revenue as well as reducing costs. He emphasized that while cost savings are a clear early benefit, the market pull will be much stronger once AI enables companies to generate more money or deliver better outcomes for customers. If AI is just a cost-cutting tool, the ceiling for adoption is low. If it doubles your revenue, the possibilities are endless.
Mr. Haber provided a compelling example from the A16Z portfolio. In plaintiffs law, Eve uses legal AI to automate drafting and discovery. Because plaintiffs’ attorneys work on a contingency basis (they are only paid if they win), AI’s ability to automate a large number of tasks will allow plaintiffs’ attorneys to take on more cases and prioritize their efforts on high-value cases, directly enhancing their revenue model. Similarly, in consumer finance, Salient's AI voice agents handle collections and compliance. While efficiency is a key factor, the key finding is that AI agents actually achieve better recovery rates than human agents. This directly improves a lender's core financial outcomes and creates compounding benefits. Haber emphasized that the source of these models' defensive power is often their unique outcome data: “The more cases Eve processes, the smarter and more powerful the platform becomes.” These AI applications own the end-to-end workflow and capture their own outcome data, such as case success rates or loan recovery rates, creating compounding benefits that traditional software that only tracks inputs cannot achieve. This strategic alignment of AI and core financial incentives is what differentiates temporary AI capabilities from permanent, defensible platforms.
