From Access to Agents - why many stop early

In the previous post, I wrote about how AI is often perceived as new, even though it is in reality based on several decades of technological development. This post is part of the AI ​​series I'm sharing this winter, where I explore what AI actually means in practice for management, organizations and value creation.

Here I take a closer look at how AI is being used in many businesses today, and why many stop earlier than they themselves think.

In the curriculum book Generative AI for Business, Shubin Yu introduces a framework that I find particularly useful for clearing this up: The five A's for applied generative AI in working life. The framework describes a spectrum of maturity levels – from simple use of AI to more integrated and autonomous systems.

Simplified, the levels can be understood as follows, with examples many will recognize in:

Access
Using general AI tools such as ChatGPT, Gemini or Copilot to write, summarize and analyze. The value is often quick and visible, but primarily individual.

Assistants
More customized assistants with role or business context, for example a sales or HR assistant who knows internal data and working methods.

Application
AI is built into specific solutions for specific tasks, such as decision support, analysis or customer dialogue in limited systems.

Automation
AI is connected to workflows across systems, so that processes are connected from start to finish - with clear efficiency and scaling gains.

Agents
More autonomous systems that can plan, prioritize and carry out tasks within defined frameworks, with clear responsibility and control.

The decisive point is not the levels themselves, but the transition between them. And this is precisely where many people stop.

In many social media AI environments, I see a strong focus on "learning many models" or mastering as many AI tools as possible. It can be useful at the Access level, but provides limited value further up the pyramid. The higher one moves, the less it is about tools - and the more about structure, integration, management and responsibility.

When AI is connected to data, processes and decisions, one goes from individual efficiency gains to organizational core competence. It is only then that AI becomes truly strategic.

Therefore, AI and strategic use and implementation are also to a small extent an IT project. It is a managerial responsibility - in line with other strategic choices related to organisation, risk and value creation.