From ERP Screens to AI Agents: What ‘Dark Software’ Means
Dark software in the enterprise is a model where core business systems lose their graphical user interfaces and exist mainly as invisible back-end services that AI agents call through APIs to execute tasks, automate workflows, and make decisions with minimal human interaction. This shift emerges as AI agents move from simple chat tools to autonomous workers that retrieve data, analyze it, and complete processes end to end. In that world, traditional ERP vendors no longer compete for human attention with dashboards and menus. Instead, they compete to be the trusted transaction and rules engines that AI-native platforms plug into. The new question is not which ERP users like most, but which back-end the company’s AI agents select for reliability, data quality, and execution speed across finance, logistics, procurement, and compliance processes.
ERP Vendors Face the ‘SaaSpocalypse’
The classic SaaS model was built around people logging into horizontal platforms and paying per user seat. As AI agents take over that role, human usage and per-seat pricing both erode. One analyst describes a looming “SaaSpocalypse”, where generic, horizontal SaaS becomes a legacy category much like on-premise software before it. If AI agents can coordinate workflows, fill forms, and move data across systems by themselves, much of the value in UI-centric ERP suites disappears. ERP vendors then face an existential choice: become AI-ready execution engines or stagnate as bloated front ends that agents circumvent. According to Younglimwon Soft Lab’s Ho Woong-ki, software providers will survive only if they are chosen by AI agents, which shifts competition from features on screens to performance, integration depth, and process intelligence behind the scenes.
AI-Native, Vertical Platforms Threaten Horizontal ERP
AI-native platforms are emerging that focus on specific industries and tasks rather than broad, horizontal workflows. These vertical AI companies do the actual work of knowledge workers: drafting contracts, flagging spending anomalies, or managing chargebacks, then charging per transaction or outcome instead of per seat. In such models, the ERP system becomes one of several dark software services that feed data and enforce rules while the AI-native layer owns the user relationship and business value. Generic workflow wrappers lose ground when agents can assemble best-of-breed APIs for each function. For legacy ERP vendors, this changes the game from selling an all-in-one platform to proving they are the richest, most reliable source of domain data and process logic for vertical AI partners. Those who fail to specialize risk becoming interchangeable plumbing in larger AI-native ecosystems.
SAP’s Vision: From Coding to Orchestrating AI Agents
SAP’s public stance shows how deep this legacy software transformation may go. CEO Christian Klein said software development is the function most exposed to AI and suggested there is a chance “no one [will be] developing software inside SAP any more” within three to four years. Instead of armies of coders, SAP expects demand for product managers, data scientists, and business experts who can design processes, define outcomes, and govern AI. Its Autonomous Enterprise and SAP Business AI Platform position ERP as the “brain of the company”, where millions of data fields, permissions, and workflows become fuel for AI agents. In this view, ERP vendors are not building screens but curating business context, quality data, and controls that agents depend on. Development becomes configuration of policy, data relationships, and exception handling rather than manual coding.
Who Wins in a World of Invisible ERP?
As ERP vendors adapt to dark software enterprise models, bolting AI on top of old interfaces will not be enough. Winners will embed AI agents into their core infrastructure so every object, rule, and event is agent-ready by design. They will expose granular APIs, event streams, and policy layers that AI-native platforms can call safely and predictably, while keeping governance, security, and audit trails intact. That means treating AI agents as the primary users and humans as occasional overseers. It also means choosing where to specialize: owning certain vertical processes deeply, or embracing a neutral role as high-reliability infrastructure for many vertical AI companies. The race is no longer to build the most colorful dashboard. It is to disappear into the stack while becoming more essential than ever to how AI runs the business.
