Prentis launched in April 2025 and is already asking investors to value it at $1 billion. That is a fast climb for a company with no public product, but the fundraising environment for AI agents is unusually loose right now.
The pitch is simple: most office workers spend a large part of their day clicking through screens, moving data between systems, and chasing documents. Prentis trains AI models to watch how those workers navigate their computers, then builds agents that can do the same tasks on their own. The company calls its model Hive-32B, and claims it costs roughly ten times less per task than what companies pay to use the large models from OpenAI or Anthropic directly.
The pricing model is the most interesting part. Prentis charges 20% of the savings it generates for customers, not a flat subscription fee. If an insurance company saves $1 million a year in claims processing costs, Prentis collects $200,000. This aligns the company's revenue directly with customer results, which is good for buyers, but it also means those contract values are estimates, not guaranteed cash.
The company's pitch deck projects an annualized run rate of $75 million by the third quarter of this year. The contracts signed so far total up to $50 million. Both numbers are performance-dependent, and Prentis itself flags that in its own materials. Investors will be pricing in some discount for that uncertainty.
The competitive picture is complicated. Anthropic spent $50 million acquiring a Seattle startup called Vercept earlier this year, shut down Vercept's product within 30 days of the deal, and absorbed the team directly into Claude. That acquisition tells you how seriously Anthropic is taking this specific problem. OpenAI is working on it too. So is Thinking Machines, the new lab run by former OpenAI president Mira Murati.
For an ordinary business operator, the relevant takeaway is this: every major AI company is now trying to build software that can operate your existing systems without needing a human to click the buttons. The practical benefit is that these agents can work inside legacy software, old databases, and clunky internal tools without requiring an expensive IT overhaul. They operate at the screen level, the same way a person does.
Prentis is specifically targeting industries that run on high-volume, rules-heavy paperwork: insurance claims, customs documentation, healthcare administration, manufacturing procurement. Those are exactly the industries where back-office staff spend the most time on repetitive tasks.
The founding team is worth noting. CEO Ritankar Das graduated from UC Berkeley at 18 with two degrees, earned a master's at Oxford, and dropped out of a Cambridge AI PhD program to start his holding company, Titan, in 2014. Other Titan companies include a telehealth startup and an autism care provider, both of which raised significant outside funding. Das has built and sold companies before.
Hoffman and Pincus are listed as co-founders but appear to be operating in an advisory and endorsement capacity rather than running the company day to day. Their involvement matters mainly as a signal to investors.
The honest assessment: Prentis is entering a very crowded space with well-resourced competitors. Its edge, if it has one, is a smaller and cheaper model combined with a pricing structure that puts the risk on Prentis rather than on the customer. If it can prove its automation actually delivers measurable savings at scale, the 20% revenue-share model could generate serious revenue. If the savings are harder to quantify than the pitch deck suggests, the numbers will look very different in six months.