Typology 4: The Dependent
(Low steering + limited understanding)
The Dependent is situated in the quadrant comprising low steering and limited understanding. It is the only typology that is explicitly identified as undesirable in both sessions. At the same time, both sessions make it clear that this position does not represent an exceptional scenario, but rather a real risk within AI-integrated professional practices.
The Dependent combines limited active AI-steering with an insufficiently in-depth understanding of the domain and processes. This creates a situation in which professionals largely follow AI output without providing sufficient direction to the process and without being able to thoroughly assess the substantive quality or implications of that output.
What characterises the Dependent?
The professional uses AI primarily via the system’s default settings, with minimal prompting, limited context and little active control over the process. Output is largely accepted as the system generates it, with only limited adjustment, refinement or substantive intervention.
At the same time, there is a lack of sufficiently in-depth domain and process understanding to correctly assess the quality, reliability or professional suitability of that output. This creates a situation in which AI not only becomes a tool, but also gradually begins to take over the thinking and decision-making process.
The fundamental problem with this approach lies not solely in the output itself, but primarily in what is missing: substantive understanding, professional judgement, conscious steering and responsibility for the final result. The professional remains heavily reliant on the system to arrive at solutions, analyses or creative output, without sufficient opportunity to independently reconstruct why certain results are correct or problematic.
What makes this position particularly complex is that it often remains difficult to detect. AI systems can generate convincing and professional-looking output, even when users possess only limited expertise. As a result, dependence can remain hidden for a long time behind operationally useful results.
How does this manifest in both sessions?
Within the IT session, this situation is most explicitly evident in the phenomenon of ‘vibe coding’: individuals without sufficient programming knowledge who, with the help of AI, develop and present applications as if they were professionally substantiated.
Participants express clear concerns about this. Although AI can produce output quickly, the underlying understanding required to maintain code securely, structure it architecturally correctly or resolve problems independently is often lacking. Responsibility for quality and reliability therefore becomes diffuse.
At the same time, the session makes it clear that this dependence is not limited to people outside the field. Even experienced developers can gradually drift towards dependent use of AI when, under the influence of time pressure, efficiency considerations or routine, AI output is scrutinised less and less critically, whilst active steering and underlying expertise decline.
Within the CMM session, the same logic appears less explicitly, but is clearly recognisable in the concerns surrounding generic content and ‘one-size-fits-all’ approaches. When professionals use AI-generated content without a clear thematic direction or a sufficiently in-depth understanding of the target audience, context or positioning, the result is output that appears technically convincing but remains largely undistinctive or meaningless in terms of content.
The increasing persuasiveness of AI output also plays a significant role here. As systems produce ever more professional-looking texts, images and campaigns, there is a growing risk that output will be accepted more readily without sufficient critical analysis or active intervention.
Why is this a structural risk?
Both sessions make it clear that the ‘Dependent’ position is not exceptional, but stems from mechanisms that are structurally present within AI-integrated professional contexts.
A first mechanism is the low barrier to entry for AI systems. Even with limited technical knowledge or domain expertise, users can now quickly generate output that looks professional and appears to be operationally useful. This easily creates the impression that in-depth understanding is becoming less necessary.
A second mechanism concerns knowledge erosion. When professionals no longer actively perform certain tasks because AI takes them over, there is a risk that underlying expertise will gradually diminish. The Dependent does not arise suddenly, but rather as the end point of a slow process in which critical thinking, analytical skills and conceptual understanding are used less and less actively.
A third mechanism is the persuasiveness of AI output itself. As AI systems communicate more fluently, credibly and professionally, it becomes more difficult to recognise limited understanding or limited control in good time. As a result, professionals may feel that they are acting critically and independently, whilst in reality they are becoming increasingly dependent on AI output to make decisions.
This makes this position particularly difficult to identify and, consequently, difficult to correct.
Relationship to the other typologies
The Dependent differs fundamentally from the other three typologies because both in-depth understanding and active steering remain limited.
The Dependent shares the relatively low level of active steering with the Gatekeeper. The difference, however, lies in the level of understanding present. Whilst the Gatekeeper ultimately still possesses sufficient expertise to analyse problems in depth and intervene in a targeted manner, the Dependent largely lacks this ability.
The Dependent shares with the Operator the risk of relying on AI output, but differs fundamentally in the degree of active steering. The Operator does attempt to actively guide AI through prompting, workflow management and iterative refinement, but this steering often relies primarily on intuition, experience and practical routines rather than on in-depth substantive understanding. In the case of the Dependent, even this active control is limited.
The Dependent thus represents the point furthest removed from the Director, where in-depth understanding and conscious AI-steering converge.
Implications for the curriculum
The ‘Dependent’ position represents the scenario that curricula explicitly seek to avoid. However, this requires more than simply teaching AI tools or operational AI skills.
The curriculum must actively focus on both dimensions that are missing from this typology: in-depth domain and process understanding, and conscious AI-steering.
In practical terms, this means that degree programmes must provide learning environments in which students not only learn to use AI, but must also visibly demonstrate that the underlying knowledge, professional reasoning and substantive understanding are genuinely in place. According to both sessions, this is precisely where one of the greatest educational challenges of AI-integrated learning lies.
Thanks to AI, students today can successfully complete assignments, produce professional-looking output and carry out complex tasks without necessarily having a correspondingly strong grasp of the underlying concepts. As a result, assessing end products alone is becoming increasingly insufficient for reliably demonstrating professional competences.
Programmes would therefore be well advised to develop explicit learning activities in which students analyse independently, justify their choices, reconstruct errors and reason professionally, even when AI is actively used within the learning process. Not because AI should be avoided, but because those underlying competences remain essential for the responsible and autonomous use of AI later on.
In addition, it is advisable for study programmes to make students aware of the risk of gradual dependence, even among professionals who initially operate as Director. Both sessions highlight how knowledge erosion, time pressure and the increasing persuasiveness of AI output can lead to a gradual decline in active reasoning, critical analysis and conscious steering. Dependence does not necessarily arise suddenly, but often as a gradual process in which professionals rely increasingly on AI output whilst underlying expertise is utilised less actively.
Ownership does not arise automatically in this context. It requires a purposeful pedagogical approach in which responsibility, critical thinking and conscious steering are systematically developed alongside technical or creative competences. From a dispositional perspective, this means that curricula must not only develop knowledge and skills, but also foster professional dispositions that ensure students continue to actively apply their professional judgement, critical reflection and sense of responsibility, even when AI offers seemingly reliable or efficient solutions.