Serval's Catalyst builds IT automations and proactive agents

Serval made its Catalyst AI agent generally available Thursday, positioning it as an administrative layer that discovers automation opportunities, builds the required workflows, and creates proactive agents that identify and remediate IT issues before employees submit tickets.

Catalyst sits above Serval's service management platform as what the company calls a "super agent." It examines ticket history, standard operating procedures, or natural-language instructions to identify repetitive work, then drafts the workflows, forms, access policies, and dashboards needed to automate it. The system generates code-backed automations in TypeScript, pulling from connected systems like Okta, Google Workspace, and Microsoft Entra.

Why it matters

As enterprise service management vendors converge on AI-assisted workflow creation, Serval's differentiation lies in making the entire automation lifecycle agentic rather than just the workflow generation step. ServiceNow, Atlassian, and Freshworks all offer similar capabilities, but Serval argues its single conversational interface compresses what typically requires multiple tools and specialists into one administrative surface. The competitive question is shifting from "who has generative AI?" to how many steps separate an observed operational problem from a production automation.

Background agents that act before tickets arrive

Serval's most distinctive feature may be its background agents, which run on schedules across connected systems to correlate signals and draft remediations without waiting for help desk requests. In one customer example, an agent correlated network incidents across two offices using switch telemetry and DHCP data, traced the issue to configuration drift, and generated a remediation workflow for administrator approval.

"Most AI agents today wait for an employee to ask a question or submit a ticket," Serval co-founder and CEO Jake Stauch told VentureBeat. "We believe the future is AI that acts before an employee ever submits a request."

Model-agnostic architecture and governance controls

Serval deliberately avoids building its own foundation model, instead using what Stauch described as models from "frontier labs" and running continuous evaluations to determine which work best for specific tasks. In a May interview with Sequoia Capital, Stauch said the company was using OpenAI models for end-user interactions and tool calling, while Anthropic's Sonnet and Opus models produced stronger results for code generation.

The company's documentation now allows administrators to supply their own OpenAI or Anthropic API keys. Serval positions its proprietary value in the harness around those models: enterprise context, integrations, generated code, permissions, and governance controls.

Catalyst inherits the permissions of the user operating it and remains scoped to that user's team workspace. Everything it builds starts as a draft, and organizations can restrict publishing privileges or require formal review before automations become active.

Early customer results and deployment options

Corporate expense firm Ramp reports that Catalyst has made workflow building 50% faster and helped extend Serval across roughly 10 teams. The company automated 600 laptop replacements, saving 150 hours. Together AI says Serval automates 95% of its just-in-time infrastructure access requests.

Serval says more than 90% of customers adopted Catalyst as their starting point for automation during beta. The system is now enabled by default for all Serval organizations.

Customers can deploy Serval as cloud SaaS, on-premises, or in their own VPC. The company's Master Services Agreement specifies that customers retain rights to their data, and Serval says it does not use customer materials to train AI models.

Serval raised $75 million in a Sequoia-led Series B in December at a $1 billion valuation, bringing total funding to approximately $127 million. The company told Reuters that revenue had grown 500% since August 2025.

These details were first reported by VentureBeat.