AI replacing developers: bold prediction or operating model reset?
The idea of AI replacing developers in enterprise software refers to claims that intelligent agents and code-generating systems will take over most coding work, shifting value from human programmers to people who can define business processes, data structures, and controls for automated software creation and maintenance. SAP’s CEO says software development is the function most exposed to AI and that there is a chance “no one [will be] developing software inside SAP any more” within three to four years. That sounds like a jobs apocalypse, but it is better read as an operating model reset than a countdown to human obsolescence. The real story is that legacy ERP vendors want to recast development as enterprise software automation driven by AI agents, and they are using provocative timelines to force customers and investors to accept that shift.
SAP’s AI strategy: from code to process and a stressed share price
When SAP talks about AI replacing developers, it is not promising magic; it is trying to reposition where its value sits. The company is tying its SAP AI strategy to the “brain of the company” narrative, arguing that ERP data and process context are the hard parts, while code can be generated by AI agents around that core. Its Autonomous Enterprise vision and Business AI Platform frame future software creation as business process design, with AI assembling the technical layer underneath. According to MarketBeat, SAP’s share price was €134.52 as of June 23, down 35.4% from €208.35 at the start of 2026. That slide, and a €2.6 billion share buyback executed at an average price of €161, make the AI story a financial argument: AI must speed releases, compress manual effort, and improve platform stickiness enough to show up in backlog growth and margins, not just in keynotes.
Legacy ERP transformation: Dark Software and AI agents as primary users
SAP is not alone. At a recent EBSC 2026 conference themed “The AI Era: What Disappears and What Remains,” Younglimwon Soft Lab’s Ho Woong-ki warned that legacy software vendors building ERP systems must evolve quickly if they want to survive. His argument is more radical than Klein’s: enterprise software will become “dark software,” akin to a dark factory, where AI agents autonomously retrieve data, analyze information, and execute business processes without human intervention. In that world, APIs are designed for AI agents, not humans, and software providers survive only if they are selected by those agents. This is enterprise software automation in which AI agents are the primary users of ERP, running HR, finance, manufacturing, and logistics in the background. The implication is clear: the customer relationship moves from the user interface to the data and ontology layer, where vendors can encode business context and causal links for AI-native workflows.
Beyond coding: AI reshapes workflows, maintenance, and licensing power
The most important part of SAP’s and Younglimwon’s story is not that AI will write code; it is that AI agents will permeate the entire software lifecycle. SAP describes a move from software creation as code production to software creation as business process design, affecting how it tests code, documents functionality, handles defects, manages security review, and assigns accountability when AI-generated output hits production. Ho adds that APIs optimized for AI agents need to help those agents make better decisions, which pushes vendors to embed more business semantics into their platforms. He also expects ERP pricing to shift from seat-based licensing tied to employee counts toward usage-based models better suited for AI agent environments. For customers, this means AI agents software development will change not only customization patterns but also who controls upgrade paths, what counts as technical debt, and how deeply they are locked into a single vendor’s data and ontology.
What enterprise customers should do now
Customers should treat talk of AI replacing developers as a warning that control points are moving, not as a reason to panic about headcount. AI-assisted development can accelerate releases, reduce manual coding effort, and make extensions easier to build, but it can also create opaque customization if governance is weak. Ordinary users will feel the impact when AI agents start taking over operational tasks in ERP systems, so organisations need stronger process design, data quality, and exception-handling skills. On the commercial side, investors are testing whether AI strengthens platforms or compresses economics by reducing the value of seats, features, and traditional customization. SAP’s upcoming quarterly results will be one test of that thesis. Meanwhile, customers entering major ERP programs or renewals should interrogate AI-native licensing, clarify which AI features are included or separately charged, and insist on transparency about how automated agents will affect customization rights and exit options.
