New challenges and trends in enterprise development
The role of the enterprise developer is shifting from writing code to managing complex AI systems. Research shows that artificial intelligence is increasingly being used for cognitive tasks. Specifically, the Work Trend Index report indicates that 49% of interactions with Microsoft 365 Copilot involved cognitive work. According to Gartner, by 2028, more than half of enterprise GenAI models will be domain-specific, requiring developers to have a deep understanding of specific business processes.
Skill evolution and integration challenges
Modern developers face the challenge of legacy ERP systems that cannot efficiently process large volumes of data for AI. Addressing this requires new competencies:
- AI system architecture design and MLOps;
- model fine-tuning and prompt engineering;
- risk management, ethics, and AI security.
The use of low-code platforms and specialized security tools helps accelerate the integration of AI components into critical infrastructure.
AI security and risk management
AI security requires new approaches. According to the OWASP LLM Top 10, prompt injection is the primary threat. To systematize protection, developers are advised to use the MITRE ATLAS framework (for threat modeling) and NIST AI RMF 1.0 (for comprehensive risk management).
An example of a successful approach is the modernization of legacy banking systems using microservices with domain-specific models. This maintains core system stability, ensures data protection, and implements human-in-the-loop control.
What is at stake for the industry
The rapid evolution of AI forces a shift in the developer's role from writing code to managing complex AI systems and domain-specific models. Legacy ERP systems will face severe integration bottlenecks, making infrastructure modernization a critical requirement for businesses to remain competitive.
Practical steps
- Acquire essential competencies in AI system architecture, MLOps, model fine-tuning, and prompt engineering.
- Address security risks like prompt injection by implementing the MITRE ATLAS framework and NIST AI RMF 1.0.
- Modernize legacy systems using microservices with domain-specific models, ensuring data protection and human-in-the-loop control.
Prepared by a Software Ukraine member. Original publication.