Agentic AI: New possibilities and human indispensability

The rapid development of artificial intelligence continues unabated. A major topic of discussion at present is AI agents, which are designed to automatically solve complex tasks involving multiple steps. They offer a glimpse into a future with many new applications for AI systems. However, Agentic AI also highlights the indispensability of humans in crucial areas.

Agentic AI refers to AI systems that do not simply respond to direct user input, but independently execute complex chains of tasks. They automatically pursue predefined goals and make independent decisions along the way. However, these systems cannot yet be clearly defined.

Examples of agentic AI from various industries

To explore the diverse possibilities of agentic AI, let's look at some specific use cases from different areas. They clearly show how agentic AI systems differ from dialogic AI systems such as ChatGPT, Gemini or Claude.

Important: These examples merely outline possibilities for the use of agentic AI systems and thus provide an outlook on a possible future. How quickly agentic AI will actually develop and these theoretical scenarios will be transferred into everyday practice remains to be seen.

Knowledge work

A legal research agent supports lawyers in complex legal cases. The agent automatically searches thousands of legal texts, court rulings and specialist literature for information relevant to the case at hand.

It not only identifies direct references, but also recognises connections between different areas of law and establishes links to similar precedents that a human lawyer might have overlooked.

The agent summarises the information found in a structured manner, points out contradictory case law and generates initial lines of argument with corresponding references.

The lawyer would need days or even weeks to do this groundwork. With the agent's preliminary work, they can now concentrate on the legal drafting, for which human judgement, legal experience and strategic intuition remain irreplaceable.

Research

A genome research agent continuously analyses data from gene sequencing experiments and compares it with global genetic databases in real time. In this way, when researching rare diseases, he independently identifies gene mutations that could correlate with the clinical picture.

They then plan and orchestrate laboratory experiments to validate these hypotheses and control robotic laboratory systems. After each experimental step, the agent automatically plans and orchestrates laboratory experiments to validate these hypotheses by controlling robotic laboratory systems and continuously adapting the experimental protocols.

He meticulously documents every step in machine-readable lab journals, ensuring perfect reproducibility. This allows scientists to focus on interpreting the results, developing new research approaches and considering the ethical implications of their discoveries. Collaboration with the AI agent significantly accelerates the research process.

Medicine

A clinical decision support agent analyses the electronic medical records of a patient with complex symptoms, including pre-existing conditions, medications, laboratory values and imaging data. It compares this information with millions of similar, anonymised cases from global medical databases and identifies rare patterns that could indicate unusual diagnoses.

While the doctor examines the patient, the agent can suggest questions in real time. After the diagnosis, the agent simulates various treatment options and predicts probabilities of success and possible side effects.

The doctor retains the final decision-making authority. The collaboration leads to more accurate diagnoses, personalised treatment plans and better treatment outcomes, while the doctor gains valuable time for important human interaction with the patient.

The status quo of agentic AI – and why humans remain indispensable

The use cases outlined above are not yet a reality, but are likely to be feasible in the foreseeable future. They all follow a specific pattern that is central to agentic AI systems:

• Review, analysis and evaluation of extensive data.
• Detection of patterns by comparing large amounts of data.
• Transfer to human experts who make decisions and finalise the AI's preliminary work.
• Taking over the parts of the work that can be automated reduces the human workload and leaves more time for interpersonal interaction.

Human skills such as genuine empathy, ethical judgement, trust-building and creative handling of completely new situations cannot be taken over by AI in the foreseeable future.

In many areas, we as a society will also make a conscious decision that humans remain indispensable – whether for ethical, social or security reasons.

Text: Falk Hedemann

The ERGO perspective: ‘Intelligence as assistance’

ERGO CDO Mark Klein also writes about this on LinkedIn: ‘It's fantastic what's possible with Agentic AI.’ He is also convinced that AI agents will become important drivers of efficiency and automation in our industry, the insurance sector.

Mark Klein continues: ‘But some people are already hyperventilating: there will be companies that are made up entirely of AI agent departments. Some are already planning job advertisements in search of AI employees (instead of human ones). It all sounds like “the machines are taking over”.’

For those who think this way, the demand for ‘humans in the loop’ must either seem outdated or like ‘fake news’ designed to conceal the true facts, Mark Klein continues: ‘For me, however, “humans in the loop” is the only approach that makes sense. Not technologically.

Technologically, almost anything is possible, and accelerated cycles mean that much more can be done much faster. But we human users have demands. Most ERGO customers still want to make phone calls! Sure, phone bots and chat bots are becoming more and more accepted. But only to get to a real person when it really matters. We trust people, not machines, even if we are adaptable.’ That's why Agentic AI is best described as follows: ’Intelligence as our assistant.’

Text: Ingo Schenk