The end of per-seat SaaS and the rise of AI-native platforms
Vertical AI platforms are AI-native software systems built for specific industries that automate knowledge work, rely on domain-specific data and expertise, and increasingly replace traditional per-seat SaaS models by charging for usage or outcomes instead of human logins. This shift is not theoretical. When AI agents replace humans as the primary users, operating in “headless” mode, the subscription math that made Salesforce-era SaaS so comforting to investors stops working. A $300 billion single-session market wipeout signaled how fragile the old expectations have become. Generic, horizontal SaaS now looks like on-premise software did in its final decade: useful, but no longer the growth engine. The winners emerging from this reset are AI-native platforms built from the ground up to automate actions, not simply connect workflows. They target the $2 trillion white-collar services market and tap labor budgets instead of IT line items. In this world, charging per seat is a relic; charging per contract, per claim, or per recovered dollar becomes the norm.
Why vertical AI and domain expertise are the new software moat
Horizontal is becoming a liability. If a product is mostly a polished wrapper around a workflow that an AI agent can now handle alone, its value compresses quickly. Form builders, generic project management tools, SMB CRMs and off‑the‑shelf social schedulers fall straight into this danger zone. The defensible ground is shifting to vertical AI software: niche specialists with deep domain expertise, entrenched distribution, and proprietary data that frontier models cannot easily access. Legal, healthcare, cybersecurity, construction, financial services and defense share the same pattern: high stakes, complex regulation and heavy reliance on context and judgment. When software captures the exact workflows, terminology and compliance rules of such an industry, ripping it out means rebuilding years of corner cases and tacit knowledge, not just migrating records. “The most durable software businesses of the next decade will be built inside verticals, not across them.” The strongest contenders will be subject-matter experts that happen to run on AI-native platforms, not the other way round.
AI agents and AI-powered low-code are rewriting enterprise work
Inside these vertical platforms, AI agents are no longer experiments; they are becoming standard teammates. Coding agents such as Devin are built on infrastructure that lets them understand large codebases, validate their own work and actively support engineering teams instead of waiting for prompts. Customer operations show the same pattern: employees use AI assistants to create highly personalized customer briefings, automate administrative tasks and streamline engagement at scale. This agentic layer sits on top of a changing development stack that looks less like a single large model and more like a network of specialized models orchestrated for cost, latency and accuracy. Companies route workloads across multiple frontier and open-source models, sometimes using teams of smaller models dedicated to tasks like information gathering, response generation or error detection. The result is an AI stack that “increasingly resembles a coordinated system of agents rather than a single monolithic model.” That stack is what powers the next generation of enterprise software evolution.
Innovation is outpacing governance: the low-code shadow IT problem
While vertical AI platforms race ahead, AI-powered low-code is spreading even faster than governance can adapt. Low-code tools already solve a concrete bottleneck: business teams no longer wait months for developers to build every app or workflow, and employees can automate tasks, build their own tools and create AI-powered solutions. That freedom reduces IT backlogs and lets technical teams focus on harder problems. Spending on low-code development technologies is forecast to reach $58.2 billion by 2029, and more than 80% of new business applications are expected to be created using AI-powered low-code and no-code platforms, up from 20% in 2024. According to one research vice president at IDC, “low-code and no-code have become the primary way companies build and run AI agents at scale.” But innovation is outpacing governance: as adoption scales, tracking security, ownership and data connections becomes far harder, and governance problems emerge when organizations juggle hundreds of apps, makers and thousands of users.
Shadow IT risks and how enterprises should respond next
The tipping point comes when low-code tools shift from departmental conveniences to critical systems. Business users are no longer creating simple forms; they are wiring AI-powered agents that make decisions and touch core data. Without clear oversight, what emerges looks like shadow IT built on sanctioned platforms: the applications are visible, but accountability and documentation are missing, and organizations lose sight of how these tools connect to business-critical systems and data. Executives should assume that this wave will not slow down. The AI-native software company is already a different creature than any SaaS vendor before it. Guardrails must evolve to match: dedicated environments for experimentation, approved templates and connectors, defined paths from prototype to production, and regular reviews of which applications and agents exist across the organization. Every app should be treated as a living asset, not a one-time delivery. The next era will not be dominated by horizontal suites, but by domain-specific AI platforms plus governed low-code ecosystems. Those who act now will shape how their enterprises work; those who wait will inherit a sprawling, semi-automated shadow estate they barely understand.
