Adapting the Training of Future Software Engineers to the Challenges of AI Integration Into the Software Development Life Cycle

The swift introduction of generative AI systems into the industrial context of software development is causing a gap between industry needs and existing software engineering educational programs; henceforth, we term this the problem of curricular anticipation. According to our analysis of industry surveys and a case study involving a succession of three distinct modes of AI-based software development, there is a gradual transition in the requirements made of engineers, moving from programming itself to specification, validation, and management of artificial intelligence systems. Within this context, we develop the analogy with the concept of compilers: while natural language prompting can be seen as a higher-order abstraction layer than the code, it is important to note that the non-determinism inherent in large language models makes verification the central component of engineering practices. In light of the results, the proposed approach to addressing the identified problems involves the creation of an adaptive field of study characterized by four consistent principles of design, along with a competency model consisting of six clusters. The approach expands traditional software engineering education by incorporating competencies related to validation, orchestration, proper calibration of trust, and – as a reflection of the educational traditions at Dragomanov Ukrainian State University – mentoring.