ConveRact · solutions
AI agent governance for reviewed releases
AI agent governance gives your team a reviewable way to change customer-facing assistants. ConveRact connects saved configurations, response scenarios, real AI comparisons and explicit publishing so administrators can inspect a candidate before it becomes the active configuration.
Plan your reviewed agent release workflowFree integration. No installation charge or one-time ConveRact fee. Plan, usage, third-party, and custom development costs may apply.
Keep a version history of agent instructions
Enable release control for an existing assistant to capture its current configuration. Draft a candidate with revised instructions, capabilities, tone or language. The published agent keeps running with its current settings while you prepare the change. Enrolled agents cannot be overwritten through direct agent edits, widget identity settings or completed onboarding. Administrators retain immediate pause and resume controls.
Compare responses using your business scenarios
Write up to six scenarios with expected and forbidden phrases, including earlier conversation where useful. ConveRact evaluates both the published configuration and the candidate with the configured AI, approved knowledge retrieval and up to 40 active catalog items. Payment-verification and instruction-override scenarios are always included. Read both answers, inspect failed checks and compare recorded response latency. These evaluations do not send customer messages or execute business tools.
Publish only after current evidence and human review
Publishing requires a passing evaluation from the last 24 hours and an administrator review note. The server rejects results when the evaluated configuration, relevant source context or deployed API revision has changed. Workspaces with two administrators can lock in a separate-reviewer policy. A different administrator then publishes each candidate. All configuration releases and policy changes are recorded with their actor and time.
Restore a previous configuration when necessary
Administrators can restore a previously published agent configuration with a recorded reason. This emergency control preserves the current active or paused status and does not reverse payments, bookings or other completed actions. Knowledge, prices and platform settings retain their existing controls. Use response evaluation alongside real channel and workflow validation: a passing set of phrase checks is not a complete accuracy, voice or security benchmark.
Questions about AI agent governance
Does creating a version change the live agent?
No. Saving a candidate and evaluating its responses do not publish it. An administrator must explicitly review and publish an eligible version.
Does a response evaluation run payments or send customer messages?
No. It evaluates generated responses using saved scenarios. Operational tools, channel delivery and voice require their own validation. Evaluation requests use the workspace AI allowance.
Can our team require a second reviewer?
Yes. With two active administrators, you can require another administrator to publish candidates. Once enabled, that policy cannot be disabled for the enrolled agent. Any administrator retains audited emergency rollback and pause controls.
Is there an integration or installation fee?
ConveRact integration and installation are free, with no one-time integration fee. Subscription, usage and provider charges apply.