The fields of application for AI are growing, as are expectations: efficiency gains, competitive advantages, an effective remedy for skills shortages. AI agents are increasingly coming into focus as a solution. But how far have German companies really come? And where are the key levers for driving the AI transformation forward in a targeted manner?
While many companies have used the past few years to experiment, 2025 will be all about concrete implementation. AI will no longer be used only in IT or marketing, but will increasingly find its way into core business processes – from manufacturing and sales to customer service. At the same time, the AI transformation is facing a paradigm shift: the leap from generative AI to agentic AI.
Generative vs. Agentic AI: From content machine to team member Generative AI, such as language models GPT or Claude, specializes primarily in understanding, translating, and regenerating content—from text to images to videos. It responds to input and usually provides the first answer that comes to mind, without in-depth context analysis or planning. Agentic AI goes one crucial step further: it not only understands the situation, but also the context, plans its own actions in a targeted manner, makes autonomous decisions, and "thinks along with you." In doing so, it not only produces results, but also checks, validates, and optimizes them independently by incorporating additional information and tools. This "reasoning" process enables Agentic AI to solve problems independently and act as an autonomous player within a team – making it a promising building block for future corporate strategies.
Are German companies on the verge of autonomous AI? According to the Federal Statistical Office, almost half of large German companies and a quarter of SMEs now use artificial intelligence – and the trend is rising. Around 26 percent of companies worldwide are already working on developing their own AI agents. Against the backdrop of a shortage of skilled workers and increasing pressure to improve efficiency, 62 percent of German companies are showing increased interest in autonomous AI agents, significantly more than the international average of 52 percent.
But even though AI agents are talked about everywhere, the reality is quite different. Many companies primarily use AI as a tool or reactive assistant that processes predefined workflows. True autonomous agents that make their own decisions and act proactively are still rare.
This status quo is no coincidence, but rather the expression of a necessary stage of development. Companies must first build a solid foundation—especially the necessary knowledge and defined workflows—before they can take the next step: developing agent logic and transferring responsibility to AI systems. Only those who do their homework will be able to benefit from the efficiency and innovation potential of true AI agents in the long term.
From tool to agent – five stages on the path to autonomous AI To provide tangible and structured support for this development, we have developed a maturity model that provides a structured overview of the current and future capabilities of generative and agentic AI systems. The 5-level model for systematizing AI experiences distinguishes between tool, assistant, agent, professional, and innovator. It clearly shows that those who have not properly prepared for levels 1 and 2 will fail at level 3 when it comes to agentic AI.
Level 1: Tool – Making knowledge available In the first level, AI serves as an intelligent tool. It answers questions, generates content, or analyzes data—rule-based, reactive, and without context understanding. Its performance is based on available knowledge: the better prepared, the greater the benefit.
What companies should do now:
Level 2: Assistant – Modeling processes AI takes on certain tasks within fixed processes. It acts as a digital assistant along defined decision trees, for example in email responses, lead qualification, or report creation. Responsibility remains with humans. However, without a process structure modeled on the new possibilities, AI cannot play a productive role—assistants need clear frameworks for action. And this is precisely where Germany's decisive competitive advantage lies: not in the development of large language models, but in the deep, operational process know-how of companies.
What companies should do now:
Level 3: Agent – Supporting decision-making ability The actual agent logic begins in the third level. The AI plans and makes decisions independently in order to achieve a defined type of result. It analyzes the context, uses various tools, orchestrates processes, and dynamically adapts its approach. The system no longer thinks in terms of tasks, but in terms of intentions and outcomes.
What companies should do now:
Level 4: Professional – Delegating responsibilities In this phase, AI takes on a complete role – including all tasks, decision-making powers, and prioritization. It works independently within its area of responsibility, takes feedback into account, plans for the long term, and coordinates with other actors (human or AI). For example, an AI-based "product manager" that analyzes market data, creates feature roadmaps, plans A/B tests, and derives measures from them. Currently, this is still a pipe dream, but in the medium term, this level will also become a realistic goal.
Level 5: Innovator – AI as a creative thought leader Here, AI becomes an active innovation partner. It independently taps into new sources, develops its own hypotheses, and suggests new business models or product ideas that no one has thought of before. It performs unlimited iterations and simulations to achieve the best result, continuously learns, and links knowledge from different domains. This stage remains a prospect for the time being. However, with growing model intelligence, mature integration of human feedback loops, and new approaches to reinforcement learning (learning through trial and error), this vision is also becoming more tangible.
From AI usage to AI orchestration The development toward Agentic AI is not linear—it requires structure, clarity of objectives, and a willingness to integrate AI as a team member. Many companies today are between levels 1 and 2, with the first few venturing into level 3. The transition from tools to agents is more than just a technology upgrade: it changes roles, processes, leadership understanding, and value creation.
If you want to make a real impact with AI, you should strategically leverage the following levers:
Those who structure today can delegate tomorrow 2025 is the year when AI will be operationalized—through clearly defined workflows, structured knowledge systems, and the first agents. The use of generative AI remains important, but those who miss out on its further development into agentic AI risk falling behind. Only those who think systematically and build their own AI architecture cleanly will be able to hand over responsibility to autonomous systems later on—and thus reach the next level of digitalization.
All details about the 5-level model can be found in the white paper "Road to Agentic AI."