Agentic AI or Agentic AI is on everyone’s lips right now. The AI Congress on November 5th will also take up the topic. What exactly is “agentic AI”?
Huber: Agentic AI means systems that make decisions autonomously and can act in a real or simulated world without human intervention. Agentic AI not only processes and evaluates information, but can also, for example, set goals, develop strategies and actively influence its environment. In fact, we have seen massive developments in this topic in the last few years.
And what distinguishes agentic AI from classic AI or generative AI?
Huber: Classic AI, which has been around for several decades, can automate tasks, but is hardly interactive and usually only reacts. Rule-based systems, learning-based systems or statistical models are used for this. Generative AI based on neural networks, which has been widespread since 2022 at the latest, generates new content such as text and images and is interactive in the sense that it adapts its behavior to user requests. The best example is the chatbot ChatGPT. Finally, as mentioned, agentic AI decides and acts autonomously; it can actively interact with its environment and adapt to it.
Can you specify that?
Huber: Gladly. The example of machine maintenance can be used to clearly illustrate the differences between classic, generative and agentic AI: Classic AI allows the service life of machine components to be predicted using sensor data. Using generative AI, the technician can receive support with troubleshooting or maintenance, for example by analyzing error codes and identifying maintenance measures in dialogue with a chatbot. Agentic AI uses the predictions of classic AI, uses it to plan and coordinate maintenance operations and documents everything automatically.
What advantages does agentic AI bring to production?
Huber: With automation solutions, agentic AI can adapt to changing conditions and optimize independently instead of just executing predefined tasks. When optimizing processes, it uses continuous learning to improve processes in real time and independently of predetermined algorithms. In quality control, it can perform adaptive checks and learn from detected errors to continuously improve quality. When interacting with customers, agentic AI creates personalized, contextual responses and learns from interactions to improve service.
And what role does robotics play in all this? For agentic AI, is robotics basically the body with which the AI can interact with the physical world?
Kraus: Yes, you can say that. But agentic AI can also function without this body. And this is currently much easier to implement if it is ‘only’ about software-based agentic AI. But of course the results are different – purely data-based, and there is no physically visible action. Therefore, there are actually several advantages to using agentic AI for robotics. However, the link between agentic AI and robotics is currently still in its early stages.
How come?
Kraus: Well, robotics research is currently concerned with providing robots with a so-called “world model”, i.e. giving them a more general understanding of their environment and their task. For this purpose, many vision-language-action models, or VLAMs for short, are currently being created, which can process and evaluate sensor data from image processing, voice input and action instructions and plan an action based on this. With this foundation, robots can understand and carry out tasks. On this basis, agentic AI can be effective in the robotics context. This would allow robots to make decisions independently and without human intervention, to replan problems and ultimately develop their own solution strategy.
Und how can human-robot interaction at agentic Systems as intuitive as possible be designed?
Kraus: Natural user interfaces can do that Make the use of AI-controlled robots much easier, because then no specialist knowledge, for example in the form of a programming language for the robot is needed. This simplifies the interaction, offers a lower barrier to entry and increases efficiency when dealing with robots. Technologically, there are currently several approaches for better interaction. To the natural ones User interfaces include voice or gesture control. There are also feedback systems that provide acoustic or visual feedback Make user navigation easier. Also interesting the personalization that allows robots to adapt to the Adapt user preferences and habits.
Back to Agentic AI in general: who drives is the topic particularly strong? industry or research?
Huber: At the moment it is mainly large technology companies that are driving this the topic forward. These actors invest significant resources in research and development, to agentic AI in your products and services to integrate and thereby to achieve competitive advantages. Particularly big tech companies like How Google, Microsoft and Amazon are leaders in this area they have both the financial resources and the infrastructure to create innovative solutions develop.
And where does that leave the middle class?
Huber: For successful transfer to medium-sized companies Companies need targeted research and development work. These include domain-specific, resource-saving ones Agent architectures, transparent decision-making mechanisms, controllable autonomy functions as well as open, locally operable ones Frameworks. Close integration is particularly important with existing machines and IT systems and work processes of the company. The Research can make a crucial contribution here achieve by being practical, robust and trustworthy agentic AI solutions developed, that do not require a large cloud infrastructure and provide real added value typical SME application areas like maintenance, Production planning or order processing offer.
And what does that mean? all for the topics at this year’s AI Congresss? What highlights visitors are allowed expect on November 5th?
Kraus: We could too very well-known this year companies to participate at the AI Congress win. Be there a keynote speech by Nvidia and then lectures from Audi Sport, Schunk, Intrinsic, Siemens, GFT Germany and Stieler. It was important to us that the… Company specific Insights into development and usage more agentic Give AI and how the interaction between AI and robotics looks like. These include, for example the use of physics simulations for training robots or support of AI and robotics picking processes. That promises innovation and practical lectures.
Why should industry decision-makers straight Visit the AI Congress now?
Huber: On the one hand, AI is that Technology of the hour and their developmentspeed is massive. On the other hand, we see at the same time too little AI in practice, the added value brings. This is exactly where the congress format comes in to: Companies report for companies from theirs Learningsthe hurdles and progress all about AI in industrial applications. The speakers convey concrete use cases and Experience knowledge. And of course rounds off the opportunity for personal exchange event.”