There’s a shift under way in how enterprises deploy AI. As capabilities improve, the question is no longer “Will this even work?” but “How do we control it when it does?”

At Serval, we believe the answer lies in robust, human‑in‑the‑loop approvals that gate an AI agent’s access to sensitive tools. Before an AI agent can reset a password, offboard a user, or grant access to a privileged resource, a human should review and approve that action.

We're excited to announce the first of many upgrades to Serval’s workflow approval system:

Manager approvals

In addition to naming specific users or groups as approvers, you can now route an approval to the requester’s manager. Serval looks up reporting lines in Okta or any connected identity provider.

Multi‑step approvals

Now you can add as many approval stages as you need. For example, an MFA reset might require a manager’s sign‑off followed by a security‑team review.

Approvals can be granted right from Slack or in the Serval web portal.

In a world where AI’s power grows daily, enterprises need tools that make human oversight both robust and easy to configure. With these new capabilities, Serval continues to deliver on both fronts. Book a demo to learn more.