3 Startups · $25M+ · Agentic AI OS + PE Due Diligence + AI Agent Security

What they're building

Eragon is betting that enterprise software as we know it - buttons, dashboards, pull-down menus - is about to be replaced by a prompt. The product is an agentic AI operating system for businesses: instead of logging into Salesforce to check your pipeline, or Tableau to pull a report, or Jira to assign a task, you ask Eragon. It post-trains open-source models (Qwen, Kimi) on a company's own data and infrastructure, and deploys them inside the company's cloud environment, so the model weights and data never leave the client's servers. The demo involves onboarding a new enterprise customer entirely by prompt: Eragon automatically provisions credentials, spins up a new instance, and starts an onboarding workflow, all from a single natural language instruction.

Why this matters

Josh Sirota spent years in go-to-market at Oracle and Salesforce implementing enterprise software for large organisations, and his core insight is that the interface itself has become the bottleneck. The average knowledge worker touches six to ten different SaaS tools per day, each with its own learning curve, data model, and UI. Eragon's thesis is that AI is finally capable enough to collapse that complexity into a conversational layer that knows the company's data and can act on it. The company-data-stays-in-your-cloud angle is strategically important: Nvidia's Jensen Huang made nearly the same argument at GTC last week, describing how "every SaaS company will become Agentic as a Service" and how enterprise AI will require local, bespoke deployment. That validation from the most important infrastructure company in AI is a significant tailwind. Eragon already has customers including Corgi, an insurance startup that raised $180M post-YC, whose CEO called it "the best applied AI for enterprise in the market" which is a credible reference from a credible source, not a PR puff quote.