No-Code erforderlich

: Einsatzmöglichkeiten der No-Code/Low-Code-Plattformen und der generativen KI im Software-Lebenszyklus

Abstract

The development of software is a complex and time-consuming task. Requirements must be recorded correctly, the software architecture should reflect the customer's needs as accurately as possible and the implementation of the requirements in programme code should be as efficient as possible, given that this generates a lot of costs for the enterprises. At the same time, the created application still needs to be tested and maintained. In this context, the demand for software solutions exceeds the number of available number of developers. No-code/low-code platforms aim to democratise programming and generative AI to automate repetitive tasks. McKinsey therefore proposes a new approach to development in which developers replace programming with drag-and-drop tools and let the AI perform automated tests. In addition, coding models such as Codex and GitHub Copilot should provide direct support and act as coarchitects. Although these technologies seem promising, it requires a deep dive to identify the real impact of the technologies on the traditional software lifecycle. This will also help to better define the future role of developers. This thesis therefore analyzed the influence of no-code/low-code platforms and generative AI in the software life cycle. First, a systematic literature review was conducted to analyze the current state of knowledge about the technologies and the influences on the phases and activities within the life cycle. In this way, the technologies could also be defined and described. This made it possible to answer the research question: „How do no-code/low-code platforms and generative AI technologies influence the tasks within the phases of the software life cycle?” The summarized results show that no-code/low-code platforms offer a suitable alternative for non-programmers but are unsuitable for more complex projects. They can therefore be used in part for rapid prototyping but cannot yet deliver the same results as customized development due to the prefabricated models. In addition, large companies are very skeptical about no-code/low-code platforms and generative AI, which is why long decision-making processes are associated with them. AI tools such as ChatGPT, CodeBART and GitHub Copilot show automation potential in various phases of the software life cycle. Starting with requirements capture through to maintenance. Simple to medium tasks can be automated with AI, but current models are not yet able to function without the technical understanding and algorithmic thinking of an experienced developer. Although the literature suggests many possible applications, the technologies are very rarely used for simple tasks. The experts are of the opinion that they will be supported by tools, but it is very unlikely that they will be replaced by them. Instead, the quality of future programmers is called into question.

Date of Award	2024
Original language	German (Austria)
Supervisor	Thomas Schwaiger (Supervisor)