Prentis, new AI lab co-founded by Reid Hoffman, Mark Pincus in talks to raise $100M

The Silicon Valley venture landscape is witnessing a pivot from the era of large language models (LLMs) to the age of large action models (LAMs). At the center of this shift is Prentis, a nascent artificial intelligence lab co-founded by industry titans Reid Hoffman and Mark Pincus. Currently in discussions to secure roughly $100 million in seed funding, the startup represents a high-stakes bet on the "agentic" turn in AI technology. While the first wave of generative AI focused on creative output and chat interfaces, Prentis is positioning itself to master the more pragmatic, if unglamorous, field of routine task automation.

The pedigree of Prentis’s leadership provides substantial weight to its ambitious fundraising goals. Reid Hoffman, a co-founder of LinkedIn and a former board member at OpenAI, has been a central architect of the current AI boom through his involvement with Inflection AI and Greylock Partners. Mark Pincus, the founder of Zynga, brings a deep understanding of consumer behavior and platform scaling. Their partnership signals a move toward "applied AI"—software that doesn't just suggest text or code but actively navigates the digital world on behalf of the user. This follows a broader trend where senior tech figures are moving away from general-purpose research to build focused, execution-oriented infrastructure.

At its technical core, Prentis is focused on automating browser-based workflows. The startup’s philosophy is rooted in the belief that the next frontier of productivity lies in teaching AI to interact with existing software hierarchies—clicking buttons, filling forms, and navigating complex web interfaces—rather than simply generating new software scripts. This approach treats the internet as an operating system. By developing agents capable of executing multi-step processes across different web applications, Prentis aims to eliminate the friction of manual data entry and repetitive administrative labor that still defines much of the modern white-collar workday.

The business mechanics of such a venture are designed to address a growing bottleneck in the enterprise sector. While AI has become proficient at writing code, the implementation of that code often creates more complexity for non-technical users. Prentis seeks to invert this by making the browser the primary site of automation. If successful, their technology could function as an intelligent layer that sits atop legacy systems, allowing companies to automate tasks without needing to rebuild their entire IT stack or wait for API integrations that may never materialize. This "overlay" strategy could democratize automation, moving it from the hands of software engineers into the hands of general knowledge workers.

Industrial implications for this shift are profound, particularly for the established leaders in Robotic Process Automation (RPA). Companies like UiPath and Blue Prism have long dominated the automation space, but their solutions often require rigid programming and fragile scripts that break when a website’s UI changes. Prentis represents a new breed of "vision-based" or "semantic" agents that use LLMs to understand the context of a webpage. This creates a competitive threat to traditional SaaS models; if an AI agent can perform tasks across any platform, the specific "moat" of an individual software application becomes less relevant than the intelligence of the agent navigating it.

As Prentis moves toward closing its inaugural nine-figure round, the industry will be watching for a functional prototype or a "stealth" launch of its agentic framework. The primary challenge will be reliability; the "hallucinations" that plague text-generating AI are catastrophic when applied to financial transactions or data migrations. Investors and competitors alike are waiting to see if Hoffman and Pincus can solve the persistence and safety issues that currently prevent autonomous agents from achieving mass adoption. If Prentis can prove that its agents can operate with near-perfect accuracy, the startup could fundamentally redefine the relationship between humans and their computer screens.