Interview: From GenAI to Agentic AI – What Will Make the Difference in 2026

Sarina Hermann

GenAI can deliver a significant boost in productivity. However, the benefits often remain limited to isolated instances: a quick text, a concise summary, or a solid draft. Business impact only materializes when such individual results are embedded in processes. And that’s where we see a major gap: organizations often have a vision, but implementation remains fragmented.

Sarina Hermann

In many cases, GenAI is reactive: I ask a question and get a result. Agentic AI, on the other hand, is goal-oriented: The system works toward a solution, can plan subtasks, prepare decisions, and—within defined parameters—also trigger actions. This means we’re no longer just talking about content or answers, but about process steps that take place within systems: classifying tickets, verifying orders, delivering content, enriching data, and initiating workflows.

Sarina Hermann

Autonomy only works if the foundation is right. Ultimately, an agent executes whatever processes, data, and rules dictate. If processes are unclear or data quality is inconsistent, you’re not automating successful outcomes—you’re potentially just creating chaos. In conversations with our customers, we often notice that processes do not yet have the maturity required to cleanly implement agent-based orchestration. And without high-quality data and clear processes, the success of AI remains limited.

Sarina Hermann

There are three classic statements I hear particularly often. First: “We just need the right model.” In reality, however, it requires a combination of data access, role-based access control, a security strategy, and monitoring. Second: “Let’s start with one agent—the rest will fall into place.” Agentic AI is rarely a standalone system. Rather, it’s about orchestrating specialized agents that work together and must be seamlessly integrated into existing platforms. And third: “Governance can come later.” In reality, a lack of transparency, compliance concerns, and missing control mechanisms are among the biggest obstacles. In our Digital Excellence Outlook 2026, which we published in collaboration with the Handelsblatt Research Institute, governance is identified as one of the top challenges for sustainable AI deployment. Those who do not address this issue from the outset will get stuck in pilot projects.

Sarina Hermann

Business units need, above all, a clear understanding of the benefits: Where can we save time? How do we increase conversion rates? What improves our service quality? From an IT perspective, security, control, and scalability are crucial—achieved through seamless integrations, access controls, and traceability. Above all, however, it is essential to clarify which decisions AI agents are allowed to make—and which they are not. What often helps, in our experience: don’t start with the technology question, but with a clear process goal. Then it is possible to specifically assess which agent capabilities are necessary—and which data, systems, and rules must be in place to support them.

Sarina Hermann

We see ourselves as a partner that builds bridges: from strategy and the identification of meaningful AI use cases, through the establishment of a solid data and process foundation, to implementation, governance, and change management. This requires a deliberate turning point, including employee training, infrastructure adjustments, and data preparation. We help companies move beyond mere GenAI experiments and achieve true AI maturity.