Is artificial intelligence with large language models a threat to individual jobs or industries?
Anthropic has introduced a specialised large language model designed to support financial analysts. Another model can assist with legal matters and, for example, draft legal documents independently. In marketing, too, many tasks, from marketing strategy to advertising slogans, are being developed by AI. This could lead to marketing specialists becoming less important. AI is also intended to identify previously unknown vulnerabilities in open source software libraries. These initial experiences have given rise to fears that the products of traditional software manufacturers will be replaced by AI solutions at a fraction of the cost.
Is this the broad experience in the economy, or is it the marketing activities of AI providers who have invested billions of dollars in their systems?
Efficiency gains and automation potential
Experience shows that there are areas in which the knowledge learned by AI offers significant efficiency gains for individual occupational groups. For example, with the help of LLMs, language translations can be carried out in such a way that only corrections are necessary. This can lead to massive staff reductions in this area.
Similarly, chatbots can be built to answer standard questions from customers or prospects. The same applies to automated appointment scheduling with customers. This relieves the burden on service personnel or even makes them redundant.
The search for product offers can be made more targeted and efficient through precise specifications. An LLM can prepare summaries of texts. They are also able to compare text content or search for specific content in documents, even with vague descriptions. Unstructured (text) data can be transformed into structured formats that can be processed automatically. This further reduces the amount of input required and overcomes media breaks. Standard software will continue to provide process support and automation.
Limitations, post-processing and ‘human in the loop’
Meeting notes can be summarised and created by ‘listening’ AI systems. Software development can also be made more efficient in some areas. For tasks that are not highly complex, code can be created that then usually needs to be post-processed. Finding programming errors is supported, as is the documentation of the source code. In these cases, however, an expert is still needed to check, correct and supplement the proposed results. So it is not possible without ‘human in the loop’, the person who checks the AI results. The question is how time-consuming this manual follow-up work is. Our own experience with more complex tasks shows that this can be more time-consuming than solving the problem manually without AI, as there are still many errors even with AI support. The subject area and the available knowledge base of the AI model are relevant in each case.
Agent-based AI allows the solution of distinct subtasks to be linked together, thus supporting more complex tasks.
Impact on qualifications and the labour market
As a consequence, it can be said that simple tasks involving the reproduction or summarisation of knowledge and the generation of similar patterns can already be easily automated today. This may not lead to the elimination of certain tasks and thus personnel in the first step, but the tasks will be significantly accelerated. Employees without in-depth knowledge and broad experience who are not trained to make decisions will be reduced.
In cases where the results need to be reviewed more comprehensively, the people using AI will need comprehensive specialist knowledge. This also applies, for example, to how queries are submitted to such a system using so-called prompting. In fields where AI achieves efficiency gains by providing basic solutions and contributes to overall time savings, even in personnel reviews and further processing, personnel will be particularly valuable in the future. However, it should also be borne in mind that AI will not be able to handle a wide range of tasks satisfactorily at first.
This also applies to credit management. AI-supported solutions provide information and recommendations for action that must be assessed by personnel. This means that employees with experience, in-depth expertise and decision-making skills will continue to be desperately needed and are indispensable in our companies.