The path from "AI for me" to "AI for us"

AI for me Was Just the Beginning

Most AI tools today are still, at their core, single-player solutions. My account. My chat. My prompts. My files. My results. That works brilliantly for personal productivity. But a company is not a single user.

And that's exactly why I believe we're standing at the beginning of a new generation of enterprise software.

I've Seen This Shift Before

Before we founded nuwacom, I built dirico, a collaboration platform for large enterprises. In the process, I learned something crucial:

The actual function of a piece of software is often the easiest part.

It gets harder once 1,000 or 10,000 people need to work with a solution. Because then it's about roles and permissions, teams, approvals, versions, integrations, security, and governance. And above all, about collaboration. I'm watching this exact transition play out again right now with AI.

A Chat Archive Isn't a Workplace Yet

Of course, the major AI vendors have massively expanded their products. There are enterprise tiers, SSO, user management, and an ever-growing set of governance features. But a company doesn't just need 10,000 personal AI accounts. It needs shared AI infrastructure.

Here's a simple example: every day, we produce valuable work with artificial intelligence. Analyses, concepts, research, presentations, documents, or apps. But where does all of it end up? Mostly in a list of past chats. Maybe organized into projects or folders, at best.

For an individual, that might be enough. For an organization, it isn't. Companies need chats to become workspaces. People need to be able to work on projects together. To master their daily work, they need to share content, organize tasks, build on results, and assign ownership.

That doesn't require a second Jira. A simple project structure is often enough - more Trello than SAP. But the projects need to be directly connected to AI.

Memory Sits at the Center

At the same time, AI can't start from zero with every conversation. It should know my context. My work. My projects. My preferences. My past decisions.

That's what creates a personal memory. Alongside it sits the organizational memory: knowledge, documents, decisions, projects, processes, and experience. Always with the right permissions attached, of course. On top of that comes real-time access to the systems people work with every day: CRM, ERP, Microsoft 365, email, databases, and specialized line-of-business applications.

Only by connecting this data does more context emerge with every use. The system becomes more valuable, for each individual employee and for the entire organization, the more it's used.

A Meeting Illustrates This Pretty Well

Many companies already use AI solutions for meeting transcription today. The meeting gets recorded. AI produces a transcript and meeting minutes. That's useful. But the process usually ends right there. The minutes become the result. When really, they should be the starting point.

Almost every meeting produces tasks: prepare a proposal, update an analysis, coordinate a follow-up call. Shouldn't these action items land directly in shared task management? And why should it stop there? A task can be connected to an agent, a skill, or a workflow.

The research agent handles the competitive analysis. The proposal gets prepared based on CRM data and organizational knowledge. Scheduling runs automatically. Whatever AI can handle, it handles. Wherever a decision is required, a human takes over.

That, to me, perfectly illustrates the generational shift ahead: the first generation of AI generates output. The next generation organizes and gets work done.

And Suddenly Employees Are Building Their Own Software

There's a second development on top of this. Going forward, we won't just buy business software anymore. We'll increasingly build it ourselves. Vibe coding makes it possible to create small business applications in almost no time.

HR needs a tool to prepare for interviews? Build it.

Sales needs an app to review proposals? Build it.

Marketing wants briefings analyzed automatically? Build it.

And then: share it.

With the team, the department, or the entire company.

This is how an internal marketplace for apps, agents, skills, and workflows emerges. It fundamentally changes the role of enterprise software. Employees no longer just use software, they create it.

That's Why I Call It an AI Operating System

"AI operating system" is a big term. But it describes pretty accurately what's emerging here: a shared software layer between AI and the organization. Personal memory and organizational memory sit at the center. Around them, knowledge, collaboration, projects, tasks, agents, apps, and workflows take shape.

The AI operating system is the central platform. It connects to existing enterprise systems and pulls in current information whenever needed. And it brings along everything enterprise software has always required: identities, permissions, governance, security, administration, and monitoring.

Then there's cost control. AI costs no longer accrue per user alone. Models, tokens, agents, and workflows all generate variable costs. Companies need to know who's consuming what, what individual processes cost, and what value they create.

From AI for Me to AI for Us

The last few years have shown what happens when you give every person a personal AI assistant. But for me, the more interesting phase is only just beginning. What happens when AI knows our knowledge, can access our systems, and people and agents get work done together?

When meetings turn directly into tasks? When employees build their own business apps and share them with colleagues? And when all of this works within the rules and structures of an organization? Then we're no longer talking about a better AI assistant.

We're talking about a new category of enterprise software.

The first generation was AI for me.

The next generation is AI for us.

The AI operating system.