We’ve taken another step forward in coexisting with a relatively new inhabitant of Earth: artificial intelligence. What initially sparked mostly amazement – sometimes euphoric, sometimes suspicious, almost like encountering a newly discovered species at the zoo – has now given way to a more sober yet equally exciting question: How do we actually shape our relationship with this new counterpart?
Because one thing has become clear to most people: AI isn’t going away. This makes it all the more important to view it not merely as a tool, but as part of a new dynamic. Anyone working with AI must be able to assess what it can do, where it goes wrong, how it is influenced, and how easily we ourselves can be influenced.
A look at the talks and debates in the re:publica 2026 program makes it clear: fascination with the technology alone is no longer enough. The questions have become more fundamental. Who controls AI? Who benefits from it? Whom does it make dependent? And what are the social costs when machines increasingly influence our work, decision-making, and interpersonal spaces?
So the smoke surrounding the technology has cleared somewhat. Now we’re getting down to the nitty-gritty. We’re beginning to understand that AI isn’t the only thing shaped and manipulable by training data. We humans are too – through emotions, trust, routines, and closeness. AI as a machine for relationships and dependency: that’s creepy. On the one hand.
On the other hand, this is precisely what prompts us to reframe a crucial question: How do we humans actually want to collaborate with artificial intelligence? And almost more importantly: How do we live up to the responsibility that, in the end, it is still we humans who decide what machines are allowed to do on our behalf?
To do this, we must not only better understand AI, but also ourselves. We must recognize when we judge, when we trust, when we follow, and when we are being manipulated. For example, when we immediately believe a familiar voice without even noticing that it has long since been replicated or rearranged by an AI.
This is exactly where two teams from the HPI d-school came in. Together with the Potsdam-based startup VoiceChain, they used Design Thinking to develop approaches for maintaining trust in the authenticity of voices even in an age of synthetic audio content. The question behind this is highly relevant: How do we maintain a sense of orientation when even the familiar becomes technically reproducible?
Samuel Tschepe’s new white paper takes an even more fundamental approach: “The Agency Continuum – A Strategic Framework for Orchestrating Human and Artificial Intelligence”. In it, he examines, using five scenarios, how agency – that is, the power to act – can be consciously distributed between humans and machines. The central shift here is remarkable: humans are no longer merely users or decision-makers, but become orchestrators. Their task is to consciously shape the interplay between human and artificial intelligence: sometimes steering, sometimes delegating, sometimes reviewing, and sometimes operating entirely without AI.
This requires that we analyze work and decision-making situations very carefully. What do we really need from AI in a specific situation? Support? Relief? A change of perspective? Automation? Or perhaps a deliberate pause? This is no easy task – after all, we often find it difficult even in interpersonal interactions to clearly articulate what kind of support we actually need. In this interview, Samuel Tschepe explains what led him to these reflections, what patterns he has observed in discussions with seminar participants, and why Human-AI-Agency demands one thing above all else: greater awareness of when we lead, when we follow, and when we would be better off pausing to reflect.
