What Prentis Builds
Prentis develops computer-use agents — AI systems that can operate desktop software the way a human would, clicking through interfaces, filling forms, and processing documents. The initial focus is on high-volume office workflows:
- Insurance claims processing: Automating the review and processing of claims across multiple systems
- Customs refunds: Handling the complex paperwork and cross-referencing required for trade compliance
- Enterprise document workflows: Navigating legacy software systems that lack APIs
The Hive-32B Model
Prentis built its own 32-billion-parameter model called Hive-32B specifically optimized for computer-use tasks. The company claims it outperforms frontier models from OpenAI, Anthropic, and Google on targeted office automation benchmarks while costing approximately one-tenth as much per task.
This cost advantage is significant: enterprise automation at scale requires processing thousands or millions of tasks, and frontier model pricing can make large deployments prohibitively expensive. A purpose-built model that achieves similar quality at a fraction of the cost opens up use cases that weren’t economically viable before.
Business Traction
Despite launching only three months ago, Prentis has reportedly booked up to $50 million in contracts across healthcare, manufacturing, and retail. The company targets industries with high-volume, rules-based workflows where computer-use agents can deliver immediate ROI.
The Founding Team
- Ritankar Das (CEO): Co-founder and lead
- Reid Hoffman: LinkedIn co-founder, Greylock partner, and prominent AI investor
- Marc Pincus: Zynga co-founder and serial entrepreneur
The combination of enterprise AI expertise and Silicon Valley startup experience positions Prentis to compete in the rapidly growing enterprise automation market.