Serval's $1B bet: AI agent builds enterprise workflows from tickets

Serval launches AI that writes enterprise automation

Serval has released Catalyst, an AI agent that analyzes a company's IT support ticket history to identify recurring problems, then automatically generates the code and workflows needed to automate them. The approach targets a longstanding friction point in enterprise IT: turning simple business rules into working automation can take months of manual configuration work.

CEO and cofounder Jake Stauch described a CFO giving an IT leader a straightforward expense approval rule that required two months and hundreds of steps in Okta Workflows to implement. Catalyst examines what problems users are raising, checks what the connected systems' APIs allow, and builds the automation—then hands it to an administrator for review and approval before anything goes live.

The $1 billion startup, backed by $127 million from Sequoia, Redpoint, and General Catalyst, is positioning itself as a direct replacement for ServiceNow rather than a complementary tool or downmarket alternative. Stauch claims Serval has already begun taking Fortune 500 customers away from the dominant enterprise service management platform.

Why it matters

Serval's architecture represents a bet that AI-native platforms built around code generation can outmaneuver incumbents carrying decades of technical debt. If the thesis holds, enterprise software markets could see the same generational shift that occurred when cloud platforms displaced on-premise systems. The company sits at the intersection of two trends: AI agents that can write production code, and enterprise buyers frustrated by the gap between purchased AI products and actual deployments.

The technical approach

Serval uses a two-agent architecture. One agent creates tools and automations; another uses those tools to resolve employee requests. The separation gives administrators a clearer boundary between what AI can build and what runs in production. The platform primarily uses models from OpenAI and Anthropic through zero-retention endpoints.

Stauch compared Catalyst to Cursor and Claude Code for software engineers, calling his product "code gen for everything that's not software engineering." He acknowledged that foundation model companies could generate the code needed for a password-reset script, but argued that production enterprise workflows require security controls, approval routing, and permissions that model providers are unlikely to build because their business incentives point toward products where the model itself supplies most of the value.

ServiceNow's numbers tell a different story

ServiceNow reported $3.88 billion in subscription revenue for Q2 2026, up 24.5% year over year, with its AI business crossing $1 billion in annual contract value. The number of customers with agentic AI in production increased ninefold in nine months. Stauch countered that Serval hears from customers that less than 10% of the ServiceNow AI products they've purchased have been deployed, claiming ServiceNow is "pushing them to reprioritize the budget so that it can show AI revenue." ServiceNow did not respond to a request for comment.

Mike Leone, vice president and principal analyst at Moor Insights & Strategy, noted that nearly every enterprise vendor sold AI into 2025 budgets faster than customers could staff the rollouts. He called the criticism fair but cautioned that ServiceNow's 20 years of accumulated enterprise context—understanding processes, approval chains, and system dependencies—represents a significant moat that code generation alone cannot cross.

The displacement challenge

Serval raised a $47 million Series A in October 2025, followed by a $75 million Series B led by Sequoia. The company claims revenue increased roughly 500% between the two rounds and that customers automate more than 50% of tickets. The startup has not publicly disclosed exact ARR, profitability, or detailed customer counts.

Leone said that for a startup to genuinely displace ServiceNow, it needs named enterprise references outside the cloud-native cohort, a real answer for the configuration database and change management, and someone to own the generated code once there are thousands of workflows and a regulator asking who signed off. He credited Serval for building access management into the same product as ticketing, noting ServiceNow had to acquire that capability.

Stauch's long-term vision extends beyond IT tickets to allowing employees to automate their own work with personal agents. That could position Serval as an enterprise control layer sitting between AI agents and enterprise systems—a potentially more defensible position than a conventional ITSM platform.

Details were first reported by Victor Dey in Forbes.