For the last few years, the AI race was about scale: bigger models, more compute, scrapping more of the web to train models. That well is running dry. Public web data is largely exhausted and tech companies face endless copyright lawsuits from authors, artists, and news outlets for scrapping it unethically. This data is also a poor source for the kind of work enterprises actually want AI to do. An agent that has only ever seen public blog posts and Reddit threads has never seen how a real company negotiates a contract, navigates an incident, or builds a financial model. The next phase of AI progress in enterprises depends on data that has never been public at all: enterprise data.

Google just proved it

In August 2026, Google won a bankruptcy auction for the internal data of Spirit Airlines: about 100 million emails, 500 million Microsoft Teams messages, 30 million lines of proprietary software code, and internal operational records.

To agents, these millions of emails and messages are a detailed record of how people at a real company escalate problems, hand off work, and make decisions. That is the data frontier labs need to build agents that can operate autonomously inside a business, not just answer questions.

Google is not alone: earlier this year, Mercor began offering failed startups compensation for their internal Slack and GitHub history, for the same reason. These two cases both point at a real gap: there is no compliant, scalable way to unlock and access company data at scale.

This year, Redpine is working with multiple AI companies on company data, generating revenue back to right holders across all fields we're working with.

Redpine is built to close this gap

Redpine is the licensed data layer for AI. We work with rights holders, proprietary dataset owners, publishers, and research institutions to license their data fairly and compliantly, make it AI ready, and accessible via our endpoints for both AI labs and vertical startups building the next generation of models and agents.

New licensed data: Enterprise data

This new data extends our core thesis into the corporate world, unlocking the massive, non-public record of internal business operations,giving AI labs the fuel they need to climb the data wall.

Enterprise Data includes:

• 50M lines of non-public programming code
• 2M business presentation slides, including from consulting/banking
• 500K financial models and analyses

What this unlocks

This data unlocks the ability to build highly reliable agents that move past chat boxes and enter autonomous business operations.

Building a coding AI system? Fine-tune it on 50 million lines of production code, complete with commit, PR, and issue history. Your agent learns not just what good code looks like, but how real teams build, review, and ship it.

Building a market analysis tool? Train it on 2 million top-tier consulting slides. It learns to deliver consulting-grade material, not a generic summary.

Building a financial modelling agent? Power it with 500 thousand financial models, analyses, and forecasts. Your agent learns to deliver tailored models for every industry.

For Redpine, this is just the beginning. We are adding more data every week, and preparing new verticals for launch soon.

Let's talk data. hello@redpine.ai