How Can IT Teams Safely Let Coding Agents Build Automations?

Jake Stauch says IT teams can let coding agents build automations when the work goes through approval, a staging cycle, and an audit trail. Serval is the AI platform for that department, so IT can extend the tools to HR, finance, people teams, and legal.

Personal agents work well for consumers, SMB, and a lot of mid-market organizations, and they fail inside a large enterprise because each person's permissions are so limited. In a lot of cases someone is not allowed to bring an agent to work. Even if they were allowed, the agent can barely read email, so it cannot reach the systems and automations people want at work.

Engineering teams adopted enterprise AI early because code review makes it safe for a 19-year-old intern to build something powerful before it reaches production. Most companies have an IT department, not a team of software engineers. Serval connects through IT with service account credentials and admin-level permissions, then lets people reach elevated access only through strict deterministic approval workflows. Because Serval is also a ticketing system, asking an agent to do a task creates access requests that get documented and logged, plus just-in-time access to run that workflow once. It is a little more annoying than the personal-agent experience of Instinct and Muse, and it is what a large enterprise, especially a regulated one, needs.