SAP CEO Predicts Traditional Coding Could Disappear Within Four Years

The programmer’s future may depend less on typing than judgment.

Walldorf, June 2026

SAP chief executive Christian Klein has predicted that artificial intelligence could eliminate traditional software development inside the German technology company within three to four years. His argument does not necessarily imply that every engineer will disappear, but that humans may stop writing most code manually. Instead, developers would define business requirements, supervise autonomous systems, verify results and correct failures produced by artificial intelligence. The forecast places one of Europe’s largest software companies at the center of the debate over how quickly automation will transform technical employment.

Klein identified software development as the corporate function most immediately affected by generative artificial intelligence. Advanced coding assistants can already translate plain-language instructions into complete functions, applications and automated workflows. As these systems improve, SAP expects engineers to spend less time managing syntax and more time deciding what software must accomplish. The change would redefine programming as a process of specification, evaluation and governance rather than continuous manual production.

The executive’s comments reflect the expansion of so-called vibe coding, a method in which users describe a desired product through natural language while artificial intelligence generates much of the underlying software. This approach allows employees without extensive knowledge of programming languages to create prototypes, automate processes and modify digital tools. The system can then test, debug and revise its own output under human supervision. What once required several specialized developers may increasingly be completed by a smaller team coordinating multiple AI agents.

SAP has particular reasons to accelerate this transition because its products manage finance, logistics, human resources, procurement and industrial operations for thousands of companies. Modifying enterprise software traditionally requires extensive technical knowledge and long implementation cycles. Artificial intelligence could allow customers to request changes conversationally instead of navigating complex menus or commissioning custom code. The commercial value would come from reducing the distance between a business need and the software capable of executing it.

Klein’s prediction also represents a significant evolution from earlier estimates that artificial intelligence would generate most code while human developers continued performing a central role. He has previously suggested that AI systems could handle approximately 80 percent of software development. His latest remarks contemplate a more radical possibility in which no employee inside SAP writes code through conventional methods. The human contribution would move upward toward architecture, data quality, security and organizational design.

This transformation is already visible across the technology industry. Major companies report that artificial intelligence produces a growing share of their new code, although engineers generally continue reviewing and approving the output. At some organizations, developers use autonomous agents to resolve errors, create features and prepare software updates while they concentrate on broader product decisions. The trend suggests that code generation is becoming an automated layer within a larger engineering process.

The disappearance of manual coding would not automatically eliminate the need for technical expertise. Artificial intelligence can produce software that appears functional while concealing security vulnerabilities, inefficient logic or incorrect assumptions. Engineers must understand systems deeply enough to identify those defects before they reach customers. In critical environments such as banking, healthcare, manufacturing and public infrastructure, an undetected error could generate consequences far beyond a failed application.

Human oversight will also remain necessary because software requirements are rarely complete or perfectly defined. Organizations contain conflicting objectives, outdated procedures and information that cannot be reduced easily to a single instruction. An AI agent may generate technically valid code that solves the wrong problem. Future developers may therefore become interpreters between business leaders, users, regulators and automated systems.

The change could place entry-level programmers under the greatest pressure. Junior positions traditionally allow new professionals to develop experience through testing, documentation, maintenance and relatively simple coding assignments. Those are precisely the activities artificial intelligence can automate most quickly. Companies may reduce recruitment for traditional beginner roles while increasing demand for professionals capable of supervising AI, integrating data and understanding specific industries.

That shift creates a training dilemma. Senior engineers acquired their judgment by spending years performing tasks that future workers may no longer receive opportunities to practice. If organizations automate the foundation of the profession, they must create new pathways through which younger employees can learn architecture, security and system behavior. Otherwise, companies could become dependent on tools they no longer possess enough human expertise to evaluate.

Klein’s forecast should also be understood as a strategic declaration from a company competing to lead corporate artificial intelligence. Predicting the end of traditional coding reinforces SAP’s message that enterprise software is moving toward autonomous operations. The company is developing AI agents capable of interpreting business context and performing tasks across several departments. Its future revenue model may increasingly depend on how frequently customers use those intelligent systems rather than on conventional software subscriptions alone.

The broader economic consequences remain uncertain. Increased productivity could allow companies to produce more software, personalize applications and modernize systems that were previously too expensive to replace. It could also reduce the number of workers required for individual projects. Employment may not disappear uniformly, but the composition of engineering teams is likely to change as routine production becomes cheaper.

Legal responsibility represents another unresolved issue. When an autonomous system generates defective code, organizations must determine whether liability belongs to the developer, the software provider, the company deploying the application or the managers who approved it. Regulators may require detailed records showing how code was generated, tested and modified. AI-produced software will therefore need governance mechanisms as rigorous as those applied to human development.

Klein’s timetable may prove overly aggressive because large organizations often adopt technology more slowly than laboratory demonstrations suggest. Legacy systems, confidential data, regulation and cybersecurity requirements can delay automation for years. Companies may also resist transferring critical operations to systems whose decisions remain difficult to explain. The technical ability to generate code does not guarantee institutional readiness to trust it.

Nevertheless, the direction of change is increasingly clear. Programming is moving away from writing every instruction manually and toward coordinating machines capable of generating, testing and revising software. The profession may survive, but its traditional image is likely to fade. The most valuable developers will not necessarily be those who type fastest, but those who can define problems precisely, judge automated output and assume responsibility when the machine is wrong.

SAP’s prediction is therefore less a declaration that humans will abandon software engineering than a warning that the old division of labor is ending. Artificial intelligence may soon write nearly all the code, but people will still decide what systems should do, which risks are acceptable and whether the result deserves to be trusted. The keyboard may lose its central place, while human judgment becomes the final layer that automation cannot safely remove.

Información que anticipa futuros. / Information that anticipates futures.