Skip to main content

​
Purpose

Agent Studio is where an administrator runs the digital workforce: configuring AI Employees, watching how they perform, and improving them under change control. Five things happen here, in a loop. Solid lines are Available today. Dotted lines are Upcoming: staged trial and pilot testing before a change governs real work.
Status. In Exterview today, Studio lives in Settings → Setup → Operations: AI Employees, Agent Skills and Scoring Logs. The Agent Catalog and Agent Lineage are Upcoming. Each section below says what is Available and what is Upcoming.

​
Build

Available. Configure how agents work for your organization: the scoring criteria, the hiring pipeline and which agent runs each stage, agent skills and knowledge, and the agent’s role for each job. No engineering work, and no blank canvas. Upcoming: composing your own AI Employees from the Agent Library, each configuration a new version with an author, a reason and a diff.

​
Deploy

Locked while in flight

Available. A job’s scoring configuration locks so criteria never change underneath candidates already in progress. Everyone assessed for the job is assessed on the same criteria.

Staged rollout

Upcoming. Move a version through trial, pilot and production, so a configuration proves itself on real work before it governs real decisions.

Two-person change control

Upcoming. A change to a live AI Employee is made by one person and approved by another, and the approval is part of the permanent record.

Rollback and suspend

Upcoming. Restore any version in one action, and suspend any AI Employee without affecting the others.

​
Monitor

Activity

Available. The Activity Log records every action on your account, by a person or by an agent.

Approvals

Available. Every write an AI Employee proposes waits on an approval card, and the answer is recorded.

Reports and schedules

Upcoming. Work that needs a person lands in the Agent Inbox.

Swarm Control Plane

Upcoming. A live view of every role run: candidates per stage, calls made, interviews booked and held, reports pending, and exceptions.

​
Evaluate

Quality

Available. Automated checks run on every scored output, and their results show in Scoring Logs. Upcoming: blocking a version that scores worse than the one in production.

Fairness

Upcoming. Outcome rates monitored across groups at cohort level, with a configuration change that moves them adversely blocked before release.

​
Improve

Upcoming. Outcomes feed back. Overrides become proposed criteria changes, and each one passes through the same versioned, approved, evaluated path before it applies. Nothing self-modifies in production.

​
What is captured

The decision trail

Available. Every score override carries a stated reason, every Decision change asks for a reason, and each score records the model and prompt that produced it. Upcoming: the Decision Room, one permanent record per decision.

The Fairness Register

Upcoming. A continuous, cohort-level view of outcome rates by stage and group, with alerts when a change moves them.
Both are described in Governance & Control.

​
FAQs

An administrator or operations owner in your organization. It is a business configuration surface, not a development environment.
A job’s locked scoring configuration does not change underneath candidates already in progress. A change applies to new work.
The configuration, AI Employees, Scoring Logs and Activity Log are Available in Exterview. The Agent Catalog, Agent Lineage, the Agent Inbox, staged rollout, two-person change control, the Swarm Control Plane and fairness gating are Upcoming.