Manage & monitor AI employee performance with these four updates

Check, control, and safely update every AI Employee you configure to ensure they're doing their job well.
Control what your AI Employees say, and know how well they're doing
Anyone configuring an AI Employee in Partner Center or Business App now has four tools for managing it:
Jobs to be done
Every AI Employee is scored against three built-in jobs, each with chat and voice variants:
Information Issue Resolved
Inquiry Answered
Lead Captured
Each job is scored from real conversations in the last 30 days and shows "On target," "Below target," or "No results were found." Tip: You can also add your own custom jobs!
Guardrails
Add rules that apply to the AI employee's replies or to the visitor's message, and customize what happens when the defined scenarios occur. There are five core types of guardrails: 
Message length
Custom rule
Personal information
Blocked words
Safety (violence, sexual content, self-harm, illegal activity, harassment and dangerous content)
AI-assisted editing
Want to change an AI Employee's behavior by improving prompts or adding capabilities, but not quite sure how? Press "Modify" and describe the change you want in plain language. The AI Employee Builder will propose an edit designed to achieve the desired outcome, and wait for your approval or further discussion before implementing any changes.
Versioning
Edit without fear of changes having undesired effects. Each update is logged in an AI employee's History. Roll back to an earlier version with one click. 
Why it matters
Before this release, monitoring and optimizing an AI employee's performance was time consuming, and often tedious. The only way to tell whether an AI Employee was doing its job was to read conversations they participated in one by one. Important behavioral directives had to be worked into the main prompt. Updates were implemented with no history logged and no way back, and changing behavior meant rewriting instructions manually.
Now you can:
Check at a glance whether each AI Chat or Voice Receptionist is resolving issues correctly, answering inquiries, and capturing leads, and use those results to see which prompts need work
Stop an AI Employee from saying things it shouldn't, without rewriting its instructions
Make changes with confidence: try performance refinements, and roll back if the changes don't have the desired result
Describe a change in plain English, such as turning a chat prompt into a voice prompt, and review it before it goes live
Who it's for
Partners who configure and manage AI Employees for their business, and their SMB clients
Agencies running AI Chat and Voice Receptionists across many accounts
Partners and SMBs with strict business and industry policies that must be respected
How to get started
Available in Partner Center and Business App. 
Review and add jobs to be done
Available for: AI Chat Receptionist and AI Voice Receptionist
Go to AI > Workforce > select "Configure" on an AI Employee
Under Jobs to be done, review the built-in jobs and their status
To add your own, select "Add job" and enter a name, when a conversation is considered, how a conversation is evaluated, and a target pass rate
Learn more
Add a guardrail
Available for: Text-based channel employees. (Guardrails can’t be applied to voice employees.)
Go to AI > Workforce > select "Configure" on an AI Employee > Profile > Guardrails
Add a guardrail and choose whether it applies to AI replies or the visitor's message
Customize what happens when the defined scenarios occur
For Safety and Custom rule guardrails, consider writing your own message for blocked replies (they use a generic refusal by default)
Learn more
Make a change with AI-assisted editing
Available for: Any AI Employee
Go to AI > Workforce > select "Configure" on an AI Employee > Modify
Start from "Improve my prompts," "Create a new capability," or "Add useful capabilities for my business category," or describe your own desired change
Review the plan, including "What Stays Exactly the Same," then select "Approve" to publish it as a new version
Manage versions
Available for: Any AI Employee
Go to AI > Workforce > select "Configure" on an AI Employee > Restore
 to view version history and restore a previous version
Learn more
FAQs
Q: Will this change how my AI Employees behave today?
A: No. Guardrails and custom jobs only apply once you add or activate them. 
Q: Is the safety guardrail on by default?
A: No. Every guardrail is opt-in, Safety included.
Q: Which AI employees can I use these features on?
A: Modifying with AI and versioning are available for all AI employees. Guardrails are available to any AI employees that communicate through text-based methods, so they are not available for AI Voice Receptionists. Jobs to be Done are available for AI Chat and Voice Receptionists.
Q: What does a job's status mean?
A: Each job is scored against real conversations from the last 30 days. "On target" and "Below target" compare its pass rate with the job's target. "No results were found" means no conversation in that window matched the job.
Q: What does a rollback restore?
A: The AI Employee's name, role, capabilities, guardrails and model settings. Triggers, knowledge sources and automations aren't part of a version, so check those after a rollback.
Q: I edited a capability and other AI Employees got a new version. Why?
A: A capability shared by several AI Employees is one configuration, so publishing a change to it creates a new version of every AI Employee that uses it. To change just one AI Employee, give it its own copy of the capability.
Q: Does AI-assisted editing change anything before I approve it?
A: No. The builder returns a plan, and nothing changes until you approve it.