Two startups at global scale without DevOps

Growth Without an Ops Headcount

Leonardo.AI renders more than 4.5 million images daily for users worldwide, and Relevance AI's autonomous agents operate around the clock across Salesforce, HubSpot, Slack, and other enterprise systems. Neither company employs a dedicated DevOps engineer. The absence is deliberate—an operational strategy born of necessity and choice.

This pattern is gaining traction across the APAC startup scene, where AI-native companies are multiplying. Australia alone counts over 1,000 such ventures, per the State of Australian Startup Funding Report, while AI accounted for nearly one-third of Singapore's new startups last year.

The Talent Gap Behind the Capital Surplus

Venture funding for AI in the region is robust: more than $1 billion flowed to Australian AI startups in 2025, and Singapore drew $8.4 billion—75% of the region's total. The bottleneck is human, not financial. A DevOps engineer in Australia commands $150K+ annually, requires months to recruit, and remains scarce. IDC reports that 60 to 80 percent of APAC organizations struggle to fill IT roles.

For a 15-person startup, the cost of adding one such hire can extend the runway by three months. Tooling expenses, on-call obligations, and the diversion of engineering effort from product development push the real cost far beyond the salary itself. An increasing number of companies are responding by selecting infrastructure that obviates the hiring need altogether. Two Sydney-based startups illustrate how that approach plays out at scale.

Relevance AI: Agents Shipping Without Supervision

Relevance AI's platform enables sales and marketing teams to deploy AI agents that handle lead qualification, customer support, and outbound workflows across their existing software stack. What distinguishes the product is the breadth of what those agents can do unattended—including creating a web experience such as a landing page, testing it, and pushing it to production. Each step runs through the Vercel REST API.

The company is running roughly 50,000 agents with no infrastructure team behind them.

Leonardo.AI: From 10-Minute Builds to Autonomous Operation

Leonardo.AI initially targeted game developers seeking custom AI-generated visual assets, then expanded to serve artists, marketers, and creative professionals. Early traction was explosive—over 100,000 sign-ups and 200,000 waitlisted users by early 2023—but the underlying infrastructure buckled under the pressure. Build times routinely stretched past 10 minutes, uncached pages took a full 60 seconds to load, and outages were frequent. Engineering effort went into babysitting servers instead of refining media quality.

The transition to managed infrastructure radically altered those metrics. Build times dropped from more than 10 minutes to roughly two. The engineers formerly occupied with provisioning, scaling, and manual oversight were freed to iterate on the company's proprietary models. With Vercel Observability delivering real-time visibility into Fluid compute performance, anomalies are flagged before users experience them. Global scaling—even through peak traffic—now proceeds without manual intervention.

"Switching to Vercel transformed our workflow at Leonardo.AI, cutting build times from 10 minutes to just 2 minutes. Vercel didn't just speed us up; it changed how we innovate."

Peter Runham, Co-Founder & CTO

The Shared Model: Platform as Ops Team

Leonardo.AI and Relevance AI operate in different domains, but their organizational designs converge. Both achieve global reach, serve millions of users, and ship updates continuously without an infrastructure unit. The platform layer performs the provisioning, scaling, and observability work that once required dedicated personnel.

This approach is emerging as the default for AI-native startups, where speed-to-market and lean headcounts are competitive necessities. The companies that scale successfully will not necessarily be those with the largest operations departments—but those whose engineering resources remain concentrated on the product.