Quick Facts
- Serval is making Catalyst generally available on August 21, 2026, enabling it by default for all customers.
- Catalyst uses background agents that run on a schedule across connected systems to detect and remediate IT issues before tickets are filed.
- Serval raised $127 million total across a $47 million Series A and $75 million Series B, reaching a $1 billion valuation in December 2025.
Serval is launching Catalyst, a super agent that sits above its AI-native service management platform and autonomously finds recurring IT problems, then builds the workflows to fix them. The company is making the product generally available Thursday and turning it on by default for existing customers.
Most IT automation tools wait for a user to submit a ticket. Catalyst does not. The agent mines ticket history, inspects connected systems, and generates workflows, access policies, and dashboards before an administrator has opened a queue.
How Background Agents Work
The most distinct feature is Serval’s roving background agents. These run on a schedule across connected systems, correlate signals from multiple data sources, and draft a remediation for an administrator to approve.
In one customer example, an agent pulled switch telemetry, DHCP data, and historical tickets to diagnose network incidents across two offices. It ruled out hardware failure and wireless interference, traced the problem to configuration drift, and generated a remediation workflow.
The workflows are code-backed. In a demonstration, CEO Jake Stauch showed Catalyst detecting connected systems including Okta, Google Workspace, and Microsoft Entra, then generating the TypeScript needed to build password-reset workflows. Administrators can add approval steps or restrict who runs the workflow.
The Architecture Argument
Stauch says the platform’s design is a direct shot at incumbents. “We’ve built our platform around code-gen agents as a fundamental primitive. It’s a fundamentally different architecture from ServiceNow, one that wasn’t possible if you started a company before 2024,” he said.
He also claims ServiceNow customers are not deploying what they buy. “We hear from our customers that they’ve bought a bunch of ServiceNow SKUs and have never been able to get them implemented,” Stauch said, adding that fewer than 10% of purchased ServiceNow AI products have been deployed at those accounts.
Mike Leone, VP and principal analyst at Moor Insights and Strategy, said TypeScript-backed workflows change the economics of enterprise automation. “TypeScript gives you a new kind of automation that you can read, version, and hand to an auditor. It can be built in hours rather than through a traditional services engagement,” Leone said.
Customer Results and Model Choices
Serval’s published results are sharp. At Verkada, the platform cut median resolution time by 90%. At Perplexity, more than 50% of IT requests are handled with no human interaction.
On the model side, Serval runs both OpenAI and Anthropic. Stauch said in a May 2026 interview that OpenAI’s GPT models perform best for end-user interactions and tool calling. Anthropic’s Sonnet and Opus models produce the strongest results for code generation, which is the core workload powering Catalyst. The company runs continuous evaluations rather than automatically upgrading every workload to the newest model release.
Company Background and Funding
Stauch and co-founder Alex McLeod, both former Verkada employees, started Serval in April 2024. The company raised a $47 million Series A led by Redpoint Ventures in October 2025, then a $75 million Series B led by Sequoia Capital in December 2025, bringing total funding to $127 million and valuation to $1 billion.
In the three months after August 2025, Serval grew revenue by 500% and more than tripled its headcount. The platform has expanded from IT automation into HR, finance, legal, security, and engineering departments.
Stauch summed up the product’s goal plainly: “We want to make it easier to automate something forever than do it manually once.”
This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.