This startup wants to make enterprise software look more like a prompt

When we talk about a revolution in user interfaces, we usually think of the transition from command lines to icons, from icons to touch screens. But Josh Sirota, founder of the Eragon startup, proposes something radically different: the complete elimination of the interface. Instead, users would communicate with software through simple questions — exactly like talking to an AI assistant. This vision is not the stuff of science fiction. Eragon has just raised 12 million dollars in seed funding at a valuation of 100 million dollars to build an AI agent-based operating system dedicated to enterprises. This is not another chatbot or automation tool. It is a fundamental transformation of how we work with corporate software.

The history of corporate IT is a history of complications. For decades, enterprises have invested in ERP, CRM, HRM systems — each with its own interfaces, logic, training requirements. An employee must learn to use SAP, then Salesforce, then Jira. It is costly, frustrating, and inefficient. Eragon wants to change this by introducing an abstraction layer — an operating system that understands all these tools and allows the user to work through natural language. It sounds like a promise we have heard many times. But this time it is backed by serious capital and a team that knows what it is doing.

When the interface becomes an obstacle instead of a tool

Modern Enterprise software is archaic. Not in a technological sense — many corporate systems are technologically advanced. But archaic in its approach to human-machine interaction. Graphical interfaces, which have dominated the last three decades, were revolutionary in their time. They allowed people without technical knowledge to use computers. But today they are an obstacle. An employee must know where to click, what menu to open, what fields to fill. This requires memorization, training, procedures.

Eragon understands that a new generation of workers does not want to learn this. They come with experience using ChatGPT, Claude, Gemini. They expect to simply ask. "What is the status of project X for client Y?" — instead of going into five different systems, searching databases, gathering information. Agent AI can do this automatically. This is not limited to simple queries. Eragon positions its system as a full-fledged operating system — which means it can manage complex workflows, make decisions, execute transactions.

Such a vision makes business sense. Employee productivity increases when they do not have to struggle with interfaces. Training time drops dramatically. Errors decrease because an AI agent will not click the wrong field or skip a step. For large organizations where thousands of employees spend hours daily on internal systems, this could mean billions in savings.

Operating system architecture for AI

To understand what Eragon is trying to build, we must first explain what "operating system for AI" means in the Enterprise context. It is not Linux or Windows. Rather, it is an abstraction layer between the user and the ecosystem of corporate applications. A traditional operating system manages hardware resources — memory, processor, disks. Eragon's operating system manages information and application resources — it integrates with SAP, Salesforce, Slack, Jira and hundreds of other tools, understands their schemas, business logic, access permissions.

A key element is API integration. Eragon must be able to communicate with every Enterprise system that already exists in an organization. This is not simple — each system has different APIs, different authentication requirements, different data models. But once this is achieved, an AI agent can work directly in these systems. It can create service tickets, update customer records, approve expenses — all through natural language.

The second element is security and governance. This is critical for Enterprise. An AI agent cannot have access to everything. It must respect user permissions, access policies, audit trails. Eragon must build a system that is as secure as traditional software, but more flexible. This technical challenge is enormous.

The third element is reliability and predictability. In consumer AI, it is acceptable that sometimes the model hallucinates or makes mistakes. In Enterprise, this is not acceptable. If an AI agent approves a million-dollar transaction based on a misunderstanding of instructions, that is a problem. Eragon must build a system that is highly reliable and can be audited.

Competition and the technological landscape

Eragon is not alone in this vision. OpenAI is working on agents for Enterprise, Anthropic is promoting Claude for business, and traditional software vendors — Salesforce, Microsoft, SAP — are all integrating AI into their platforms. But there are differences. OpenAI and Anthropic are building foundational models. Traditional vendors are building features into existing products. Eragon is building a new layer — an operating system that sits above everything.

This is potentially a stronger position. If Eragon succeeds in building a system that actually works — that allows an employee to ask "What is my revenue this month?" and get an answer from ERP, CRM and analytics tools without needing to know where that data is located — then it will have a moat. It will be hard to replace because it will be integrated with the entire customer ecosystem.

However, competition is serious. Microsoft, with enormous capital and Enterprise relationships, can integrate similar features into Azure and Office 365. Salesforce is already doing this in its CRM. Traditional vendors have the advantage of existing customer relationships. Eragon must be better, faster, more flexible — and must achieve this before large corporations decide they can do it themselves.

Practical challenges: from vision to reality

12 million dollars is serious funding for a startup, but for this type of project it is just the beginning. Building an operating system for Enterprise requires not only engineering, but also deep customer relationships, understanding of their business processes, testing in real environments. Eragon must find its first customers — organizations that are bold enough to try a completely new approach to software.

The second problem is hallucinations and errors in AI models. Modern language models are remarkably capable, but they are not perfect. They can misinterpret an instruction, provide incorrect information, or make an unexpected decision. In an Enterprise context, this is unacceptable. Eragon will have to build systems that verify AI decisions, require approval for important operations, and can explain their logic. This greatly complicates the architecture.

The third problem is training and adoption. Even if Eragon's system is perfect, employees will need to learn how to use it. "How do I ask the operating system?" — this is not intuitive for everyone. Training, documentation, and support will be required. Eragon will have to build not just the technology, but an entire implementation infrastructure.

The fourth problem is integration with legacy systems. Much Enterprise software consists of systems that have existed for decades, written in COBOL or other old technologies. Their APIs are limited or do not exist. Eragon will have to find ways to integrate with these systems — perhaps through reverse engineering, interface emulation, or building custom connectors.

Polish enterprises and the future of work

For Polish companies, this has concrete significance. Poland has many large employers — banks, manufacturers, logistics companies — that are trapped in complicated IT systems. Implementing SAP or Oracle costs millions and takes years. Employees spend hours in training. If Eragon succeeds in building a system that actually works, it could be transformative for Polish business.

But there are also risks. Poland has a strong tradition of traditional IT systems. Large companies have IT teams that are interested in maintaining the status quo. A shift to an AI-based operating system would be radical. It would require a change in thinking, new skills, acceptance that machines make decisions. This will not be easy.

Business model and path to profitability

How does Eragon make money? This is a key question for any startup. Most likely it will be a subscription model — each organization pays for access to the operating system, perhaps based on the number of users or number of transactions. This is a model that has worked for SaaS companies. But for Enterprise it may be more complex — they may require custom deployment, on-premise hosting, dedicated support.

Gross margins for such a product should be high — Software as a Service typically has 70-80% margins. But implementation costs can be high. Each customer will require custom integration, training, support. Eragon will have to find a way to scale without scaling costs — perhaps through a marketplace of partners who will do integrations.

The path to profitability will not be simple. Eragon will have to acquire many customers before revenues cover costs. 12 million dollars should be enough for 18-24 months of operations for a small team. This means Eragon will have to demonstrate significant progress — several major customers, solid traction — before seeking Series A funding.

A shift in how we think about interfaces

Regardless of whether Eragon succeeds or not, it represents an important shift in how we think about user interfaces. For decades, interface design was about making it intuitive — about getting buttons in the right place, making workflows logical. But maybe that was the wrong path. Maybe the right interface is one that completely disappears — replaced by natural conversation.

This is already visible in consumer AI. People do not think about ChatGPT's interface — they think about what they want to achieve and ask. The interface is transparent. Eragon wants to bring this to Enterprise. If it succeeds, it will mean the end of the GUI era for Enterprise software — at least for many use cases. It will mean that UI/UX engineers will work on something completely different — on how to make an AI agent more helpful, more understandable, more trustworthy.

This is a shift that makes sense. Graphical interfaces were perfect for a world where computers were rare and expensive. But now, when AI is ubiquitous, we can finally realize the vision that scientists have had for decades — computers that understand us and work for us, without needing to learn their language.