Serval has launched Catalyst, an AI agent designed for enterprise automation, making it generally available and enabling it by default for its customers.
Catalyst functions as an admin-facing "super agent" that can inspect ticket history, standard operating procedures, or natural-language instructions to identify recurring work and draft the necessary workflows, skills, forms, access policies, journeys, and dashboards needed for automation. A standout feature of Catalyst is its ability to create background agents that continuously monitor connected systems for emerging IT issues and propose solutions before an employee files a ticket.
Catalyst's approach is distinct in the rapidly converging field of AI-assisted workflow creation. Unlike ServiceNow’s Build Agent or Atlassian’s Rovo, which translate natural-language instructions into full-stack applications and Jira automation flows, respectively, Catalyst is designed as a single administrative layer capable of moving from discovering an automation opportunity to assembling multiple types of governed automation. This includes proactive agents that look for new automation opportunities.
Serval does not build its own foundation model but uses models from "frontier labs," such as OpenAI and Anthropic, evaluating them for specific tasks. OpenAI’s models are favored for end-user interactions, while Anthropic’s models perform well in code generation, crucial for Catalyst's offerings. Catalyst allows organization administrators to use their own OpenAI or Anthropic API keys, although users do not have a self-service menu to choose individual models.
Catalyst is targeted at enterprise users, particularly those involved in IT service management and operations. It is designed to help organizations automate repetitive help desk requests, create workflows, and improve system monitoring and troubleshooting.
Work implications: Catalyst could significantly augment roles in IT service management and support by automating routine tasks and preemptively resolving issues, thereby freeing up human resources for more complex problem-solving.
Originally reported by VentureBeat