# Software Is Dead: Why Eragon's $12M Round Signals an Agentic Future

> Eragon raised $12M to replace traditional enterprise software with an agentic OS, marking a shift from seat-based SaaS to prompt-first workflows.

Source: https://www.starbox.work/en/insights/ai-news/2026-03-19-19/ (Starbox AI Insights, updated 2026-03-19)

"Software is dead." That was the opening statement from Josh Sirota, founder of enterprise startup Eragon, in a recent interview with TechCrunch. In his view, buttons, dialog boxes, and dropdown menus are relics of the past. Moving forward, enterprise work will run entirely on prompts.

Founded in August 2025, Eragon announced a $12 million funding round at a $100 million valuation. The company's goal is straightforward: replace platforms like Salesforce, Snowflake, Tableau, and Jira not one by one, but by absorbing their functions into a single agentic operating system powered by large language models.

That ambition reflects a broader shift across enterprise software. At GTC 2026, analysts at Constellation Research noted that Salesforce is rolling out an "Agentic Enterprise License Agreement" (AELA)—an all-you-can-eat packaging model for AI agents. When legacy SaaS leaders begin overhauling their own pricing around agents, the direction of the industry is no longer in question.

## Moving from Software Interfaces to Natural Language

Before founding Eragon, Sirota led go-to-market initiatives at Oracle and Salesforce. That background gave him firsthand visibility into recurring enterprise software frustrations: siloed data across platforms, steep learning curves for end users, and long IT backlogs for simple configuration changes. 

Eragon's technical approach bypasses these bottlenecks by keeping enterprise data inside the customer's own servers and secure perimeter. Model weights also remain company property. The CEO of insurtech company Corgi called Eragon "the best enterprise application AI on the market," pointing to its ability to train and deploy models inside the company's private environment without sharing data with third parties.

Sirota argues that proprietary models trained on years—or decades—of internal enterprise data become core corporate assets. That position matches Nvidia's message at GTC 2026: proprietary data serves as an enterprise's most durable competitive moat in the AI era.

## SaaS Anxiety: From Seat Licenses to Task-Based Pricing

The vendors Eragon aims to displace are feeling the pressure. Atlassian's stock has dropped 84% from its 2021 peak, largely due to investor concern that AI tools will render products like Jira and Confluence redundant. According to Sapphire Ventures data, pure-play SaaS indexes fell 23% in early 2026. 

Skeptics often point to MIT research showing that 95% of enterprise AI pilots fail to achieve long-term adoption. Sirota pushed back on that interpretation, remarking that pilots stall mainly because executive leadership rarely understands what operational staff actually do throughout the day.

Constellation Research frames the transition more structurally: AI agents are shifting from point features into the primary application layer. Salesforce's move toward AELA reflects a fundamental reality—enterprises are moving away from paying per human seat and toward paying per task completed by agents. When the billing metric changes from headcount to output, the traditional SaaS business model breaks down.

## The Starbox Perspective: Prompt-First Changes the Way Work Gets Done

AI is not an incremental update to legacy SaaS; it is an architectural replacement. While early-stage platforms like Eragon are just beginning to prove this concept, Starbox AI has already deployed AI-native workflows across hundreds of industry scenarios.

Adding AI as an add-on feature inside legacy software rarely solves workflow fragmentation. At Starbox AI, more than 300 production-grade agents operate within a three-tier framework (Organization → BOX → Agent). Instead of logging into ten separate software platforms to launch a marketing campaign, teams define their objectives in natural language within a single workbench. The system orchestrates the appropriate tools, customer data, and execution templates automatically.

Moving to a prompt-first model is not simply about changing interfaces. It requires deep vertical domain knowledge, structured data management, and operational frameworks that can support daily production.

Across GTC 2026—from NemoClaw and Vera Rubin to startups like Eragon—every layer of the enterprise technology stack is being redesigned around AI agents. The question for enterprise leaders is whether their existing software architecture will be disrupted from the outside, or rebuilt intentionally from within.