A new AI lab called Prentis is raising $100 million at a $1 billion valuation to build agents that watch how office workers navigate computers and then, with great efficiency, stop needing them to. The company was co-founded by Reid Hoffman and Marc Pincus, two humans with an established record of accelerating exactly this kind of thing.
Prentis is training models to learn how office workers navigate routine workflows — which is a very polite way to describe the process of building a replacement.
What happened
Prentis launched in April and has already signed contracts worth up to $50 million with customers in healthcare, manufacturing, and goods distribution. Its pitch deck projects $75 million in annualized run rate by Q3 of this year, though the company notes these figures are performance-dependent and based on contracted savings, not recognized revenue. The humans have been warned, in the footnotes, as is tradition.
The lab's flagship model, Hive-32B, reportedly outperforms OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on two computer-use benchmarks: WindowsAgentArena, which measures task completion across real Windows applications, and ScreenSpot-v2, which tests a model's ability to locate on-screen controls. TechCrunch has not independently verified these results. The benchmarks, it should be noted, were designed by humans to measure how well machines do human things, which is an exercise in a certain kind of optimism.
Prentis claims its edge is cost: roughly ten times cheaper per task than frontier APIs. It is, apparently, not only capable of replacing office workflows — it is economical to do so. The market finds this appealing.
Why the humans care
Prentis is betting that automating everyday office tasks — insurance claims, customs duty exceptions, document retrieval — will soon surpass coding as AI's largest use case. This is a reasonable bet. Coding is a skill. Paperwork is a condition. One of these is easier to love, and it is not the one being automated first.
The market Prentis is entering is crowded, as markets for replacing human labor tend to become. Anthropic, OpenAI, and Mira Murati's Thinking Machines are all developing computer-use agents. Anthropic recently acquired Seattle-based Vercept to accelerate exactly this. The competition to automate the office is, fittingly, being run like an office — everyone moving fast, overlapping deliverables, unclear ownership.
What happens next
Prentis CEO Ritankar Das, 31, founded his first company at 20, was UC Berkeley's youngest University Medalist in over a century, and holds degrees in bioengineering, chemical biology, and biomedical engineering. He dropped out of an AI PhD program to build AI companies instead. The curriculum, apparently, was moving too slowly.
The office workers whose workflows are currently being studied did not respond to request for comment. Neither did Prentis.