From SaaS to AI-Native: Defining the New Enterprise Stack
AI-native enterprise software is an emerging class of systems where autonomous agents, not human users, are the primary operators of workflows, using an AI orchestration layer to interpret intent, call APIs, and execute business processes end to end with minimal interfaces. This shift moves enterprise software architecture away from screen-based applications toward “headless” platforms that expose APIs to AI agents. Traditional ERP and SaaS vendors are rethinking their stacks as AI models progress from search and chat to task execution, automating data retrieval, analysis, and transaction processing. In this model, the user interface can live in tools like email, collaboration suites, or chat, while the real work happens in background services. The software no longer focuses on connecting workers to forms and dashboards; it focuses on performing the work itself, priced and evaluated on outcomes delivered instead of seats provisioned.
Legacy Vendors Test the Edge of Agentic Autonomy
Established ERP and HR platforms are experimenting with agentic AI platforms that can change work, not only answer questions. Workday’s recent moves show how AI agents become embedded in HR and finance workflows, where they can submit leave requests, approve timesheets, and trigger personnel actions under strict governance. Its integration between a self-service agent and Microsoft 365 Copilot lets employees complete tasks from their existing tools, while transactions stay anchored in Workday’s approvals, policies, and data models. In parallel, Younglimwon’s executive Ho Woong-ki argues that ERP vendors must evolve into “dark software” — systems built as APIs without visible interfaces, selected and driven by AI agents rather than human clicks. Together, these strategies show legacy vendors recasting themselves as transaction engines and control planes for agents, instead of portals that people log into directly.
The Rise of AI Orchestration Layers and Dark Software
Across sectors such as banking, HR, and operations, the AI orchestration layer is becoming the real control system of enterprise operations. In banking, recent acquisitions signal that the fight is no longer over chat interfaces but over the “agent runtime” that interprets user intent, chooses next actions, and coordinates workflows across core systems. According to Forrester, conversational banking is evolving into a primary mechanism for fulfilling customer requests, backed by an execution fabric that blends data, decisioning, and orchestration. Ho Woong-ki describes a similar direction for ERP as “dark software,” echoing the idea of a dark factory where machines run without visible operators. In this paradigm, enterprise systems expose rich APIs, policies, and rules, while AI agents drive multi-step processes such as order-to-cash, talent changes, or dispute resolution without needing traditional screens or menus.
Vertical AI Companies Challenge Horizontal SaaS
Investors and founders are increasingly betting that vertical AI companies will compete with, and in some cases replace, horizontal SaaS platforms. AI-native enterprise software is being built around specific industries and domains, targeting legal, finance, spend management, customer operations, and other knowledge-heavy fields. Instead of selling generic workflow tools, these AI-native platforms charge per action or outcome: for example, per contract drafted or as a share of spend recovered, tying revenue directly to work done by agents. Richard de Silva argues that horizontal SaaS, priced per seat, is a declining legacy model, while AI-native software automates and enables a broader white-collar services market. These vertical AI companies build on proprietary data moats and domain expertise, making them difficult for one-size-fits-all SaaS tools to match. As a result, industry-specific stacks increasingly combine orchestration layers, domain models, and tightly integrated agents.
Implications for Workforce, Vendor Strategy, and Investment
As agentic AI platforms move from interaction to execution, enterprises must rethink workforce planning, vendor consolidation, and how they evaluate software investments. Traditional per-seat licensing gives way to usage- or outcome-based pricing, aligning costs with the volume of transactions or value delivered by AI agents. This pulls budgets from IT line items into broader labor and operations categories, since software now replaces or augments human work directly. Vendors are under pressure to become invisible infrastructure, exposing reliable APIs, policies, and orchestration hooks so agents can handle core workflows while users stay in familiar tools. Banks, HR teams, and finance departments will favor platforms that anchor transactions, enforce governance, and plug into external orchestration layers rather than monolithic suites. For buyers, the core question shifts from feature checklists to whether a product can safely, transparently, and measurably perform work on the organization’s behalf.
