AI Is Gaining Ground, but Still Needs OversightÂ
Artificial intelligence is demonstrating increasingly extensive capabilities. The idea that it should not replace people but rather cooperate with them and make their work easier, is an important guiding principle. Yet the question inevitably arises as to the extent to which AI can take on tasks autonomously, and to what extent people need to be involved. It should likewise be clear which types of activities must be handled by people alone and where a combination of humans and AI may be worthwhile.Â
When we look at today’s workplace, one thing that stands out is that AI has evolved from a hype topic into a question of application. It is less about whether AI is used and more about exactly how it should be used. In addition, there has been a leap from the basic use of generative AI to agentic AI, with assistants that can take over entire sequences of tasks, including data-driven decision-making.Â
It seems almost paradoxical that autonomy as a significant leap in AI development is currently also the greatest limiting factor. So when we talk about autonomy in AI, it always means a capability that works only under certain conditions. In short, even if an AI already has advanced capabilities, it needs human oversight.
What Is the Decisive Factor: Humans or AI?Â
It would be fairly easy to conclude that people still make the real decisions. But the real question is what makes the decisive difference. AI capabilities are very well received and are considered important tools that companies must have today. That is true. However, many fail to recognize that the success of AI initiatives is always tied to internal prerequisites such as sufficient AI maturity, including a suitable strategy and governance, well-founded goals, sufficient expertise, and, last but not least, a high level of process maturity are needed.
AI acts as a central force and as a catalyst that drives many things forward. Yet the essential competencies required to realize AI benefits such as more efficient processes, time savings, or greater value creation through big data analyses come from people. They make judgments based on experience, prudence, and ethics, rather than on mathematical operations that cannot illuminate the whole picture. They also drive creative and innovative solutions that bring new things to light and are independent of existing data and bodies of knowledge.
This can be illustrated clearly with the example of professional AI-supported process automation: For it to be successful, a comprehensive analysis of the strengths and weaknesses of the existing process is first required. On this basis, optimization must then take place to create the prerequisites for AI implementation.
If this does not happen, only suboptimal steps and workflows will be replicated. In other words: Even with technically successful automation, the results will leave much to be desired. AI is therefore only the final major step in such an initiative and depends on conditions created by people.Â
The Individual Areas of Work: What Is Human-Driven and What Is AI-Driven?
It is no great revelation that rule-based, routine tasks have no serious consequences, and follow clear processes are best suited to AI and automation. The more data, information, and experience are available, the more likely an AI is to be able to perform a particular task autonomously. On the other hand, tasks involving new processes, innovations, and creative solutions clearly fall under human responsibility.
Overall, a distinction must be made between tasks that AI can perform fully autonomously, partially autonomously, or non-autonomously (at most in an advisory capacity).
Areas of Work With Full AI Autonomy
Full AI autonomy can be achieved in the following areas, for example:Â
- Data processing and maintenanceÂ
- Summaries and translations
- Classifications and prioritizationsÂ
- Responding to standard inquiries
- Triggering and executing standardized workflows
- Resolving known incidents according to defined procedures
- Monitoring systems
- Scheduling and coordinating appointmentsÂ
These are all types of tasks for which AI can access familiar patterns and reliably and consistently achieve good results. It becomes even more exciting when it comes to more complex tasks that contribute more substantially to value creation. Even if AI operates only partially autonomously in the relevant area, it can still have an outstanding impact that corresponds perfectly to the ideal model of human-machine interaction.Â
Areas of Work With Partial AI Autonomy
Partial autonomy means that AI performs tasks subject to human assessment and approval, and applies in the following areas, for example:Â
- Decision-making: An AI should never make complex, important, and far-reaching decisions on its own, but it can assist in preparing them. It proves helpful by analyzing data, comparing options, and making recommendations.
- Cybersecurity support: AI detects anomalies, evaluates risks, and suggests appropriate responses.
- Responding to customer concerns: In customer service, AI can make suggestions and formulate helpful responses that a person only needs to review and, if necessary, revise.
- Software development: AI is capable of generating code, making suggestions for improvements, and simulating scenarios.
- Optimizing financial management: AI creates forecasts, detects anomalies, and calculates different budget scenarios.
- Legal support: AI applications can be used to review documents, flag potential legal risks or conflicts, find solutions to them, and compile evidence.
- Content creation: This application is probably the best known and, although many users allow AI to operate fairly autonomously here, it does not work entirely without human control. In practice, AI drafts or revises written content, translations, and concepts.
A human perspective is essential in any of these areas. It is tempting to grant AI extensive autonomy, but this leads to more frequent errors, legal violations, and suboptimal results. A good example is the well-known phenomenon of AI hallucinations, which initially appear well-founded and internally consistent, but are sometimes riddled with numerous false facts. Competent people are therefore always needed to contribute context, their wealth of experience, and cross-connections, and to assume ultimate responsibility for the results generated with the help of AI.
Human Domains
In many areas of work, AI can function at most in an advisory capacity, which is not a deficiency but an important natural boundary. In the often-cited human-machine interaction, human intelligence determines the direction, strategy, areas of action, and new approaches.
Of course, it isn’t possible to provide an exhaustive account of all the areas that remain in human hands and are, at most, supported by AI advice. It must also be noted that agentic AI can indeed solve entire workflows independently, including data- and rule-based decisions. However, current AI capabilities reach their limit precisely when it comes to creating something new for which no data foundation is available.
Thus, the most important human capabilities in the AI age—perhaps even somewhat more clearly than before—are exactly what they have always been: creativity, innovative spirit, unconventional thinking, finding new solutions, imagination and original thought, as well as charisma.
Examples Illustrating Human Domains
According to this principle, for example, people must continue to make business decisions independently not only because a human being has to take responsibility for them due to their importance. Rather, the parameters on which these decisions are based are often so complex and opaque that an AI cannot arrive at a clear result. The decisive spark comes from human experience and the ability to act on limited information.
Likewise, an AI can write content, but it cannot maintain an individual style; at most, it can imitate one. AI is certainly capable of writing long texts and even books, but it cannot appeal to readers as vividly and emotionally as a good author can.Â
In everyday work, human domains and AI limitations can be simplified into this schematic approach: People develop a particular approach, have an AI application execute it, and finally review or modify the result. Â
Tasks in ITSM: To What Extent Can AI Take Over?
In ITSM, for example, AI can diagnose incidents, recommend solutions, and prepare appropriate measures. Both generative and agentic AI can be used to classify tickets, quickly create summaries and translations, generate responses, conduct sentiment analysis, or provide comprehensive support as an AI agent—even independently resolving entire workflows (end-to-end automation).Â
AI therefore does not take over ITSM (that is not its purpose) but supports agents and is clearly subject to their control. The aim is to create the most productive interaction possible, within which people experience the greatest possible relief from many previously time-consuming routine tasks and can focus more on demanding tasks such as finding (creative) solutions.Â
AI thus delivers major time savings, productivity gains, and greater value creation in ITSM. Users experience faster responses, higher service availability, and even better service quality. AI therefore provides extremely valuable support that depends on functional interaction with people in order to bear fruit.
This is exactly what OTRS’s AI services are designed for.
Humans and AI: The Secret to a Productive Combination
The use of artificial intelligence and human workers are often regarded as opposites between which a company must choose. This idea is fundamentally wrong, because the point is to combine the two and thus operating more intelligently, efficiently, and purposefully.Â
What is true is that humans and AI have different strengths, abilities, and competencies, which argues for combined use, not separate use. Only in a few cases are people simply replaced by AI; instead, they find themselves in new types of tasks and changed areas of responsibility. In most cases, this means acting more strategically, maintaining an overview, giving AI applications meaningful instructions, and monitoring them.
When humans and AI are compared, it is less about hard trade-offs and more about harmonious interaction. AI exists to support people as extensively as possible, relieve them of unpleasant and time-consuming tasks, and help them achieve better results with less effort.