Manage
Review conversations, give work a clear owner, preserve useful contact context and improve the AI Agent from evidence—not guesswork.
Use one operating loop for daily conversation work
Start in the shared inbox, identify conversations that need attention, assign a responsible person, take the approved action and record the context the next owner needs. When a conversation exposes a knowledge or behavior gap, correct one source of truth and retest before releasing the change.
Turn conversations into controlled work
The Manage workflow connects what the customer said, what the AI Agent did and what the team must do next. Every important conversation should have a visible status, appropriate owner and usable context.
Find new messages, human requests and conversations waiting for attention.
Assign the right person or queue and make responsibility visible.
Reply, follow up, pause automation or complete the approved next step.
Preserve facts, notes, tags and the reason for any handoff.
Correct the source or rule and retest the same scenario.
Review the work that needs attention first
Use the inbox as an operating queue rather than a message archive. Prioritize conversations where delay, uncertainty or an incomplete next action creates customer or business risk.
Human attention requested
The customer asked for a person, the AI Agent escalated, or the issue is sensitive, exceptional or unclear.
Action: assign and acknowledgeNext action is incomplete
A booking, estimate, application, purchase step or handoff needs confirmation or follow-up.
Action: verify the destinationNew or important message
A contact, assigned conversation or important account requires a timely operational response.
Action: review context and respondQuality signal
A repeated question, weak answer, unusual objection or routing error may require a controlled improvement.
Action: preserve evidence and testGive each handoff one responsible owner
Assignment should clarify who is responsible for the next action, when it should happen and which queue remains the fallback if the owner is unavailable.
The message, contact record and AI Agent activity establish what happened.
Use when a specific team member must act.
Use when an operating team monitors the work.
Catch work that cannot reach the intended owner.
- The reason for assignment is understandable.
- The owner has access to the conversation and required systems.
- The expected next action and timing are clear.
- Absence, reassignment and unassigned work have a fallback.
- Completion is verified rather than assumed from assignment.
- Notifications support the queue without replacing review.
Store facts, human context and reusable labels separately
Useful contact context helps the next owner understand the customer without rereading every message. Keep structured facts consistent, reserve notes for human context and use tags for approved operational grouping.
Information should remain accurate, relevant and traceable to the conversation or approved source.
Consistent facts
Name, contact details, service interest, qualification answer or another approved structured value.
Useful for search, routing and personalizationHuman context
A concise summary, exception, promise, concern or reason a person should follow up.
Useful for continuity and judgmentReusable labels
Approved categories for workflow, source, status, audience or operational grouping.
Useful for filtering and coordinationPause the AI Agent when a person should take over
Pausing automation for one contact protects an active human conversation without changing the Agent for every other customer. Use it deliberately and document what must happen before automation resumes.
Review the request, current owner, sensitivity, exception status and customer preference.
Continue with monitoring
The request is within approved knowledge and no person is actively handling it.
Pause for this contact
A person is taking over, the customer requested one, or the case requires judgment or exception handling.
Record who owns the conversation, why the AI Agent was paused and what evidence or completed action permits a safe return to automation.
Correct the source—not just the visible reply
Keep the affected conversation, identify whether the weakness came from knowledge, instructions, routing, contact context or human process, then change one source of truth and repeat the same test.
Keep the conversation, contact state and observed result.
Locate the knowledge, behavior, routing or ownership gap.
Update one approved source or rule.
Repeat the original scenario plus a nearby variation.
Approve only after the expected behavior is verified.
Approved knowledge source and conflicting versions.
Ask the same question and one related variation.
Agent instructions, ability and escalation rule.
Test the normal path and the stop condition.
Assignment logic, team access and fallback queue.
Confirm the intended person receives usable context.
Contact fields, notes, tags and handoff summary.
Verify the next owner can continue without guessing.
Use a repeatable review cadence
Conversation management works best as a defined operating habit. Match the cadence to traffic, risk, team capacity and the response expectations approved for the deployment.
Clear attention queues
- Human requests and escalations
- Unassigned or overdue work
- Important new messages
- Bookings and next actions needing confirmation
Review quality patterns
- Repeated unanswered questions
- Weak handoffs or incomplete context
- Assignment and notification gaps
- Contacts paused without a clear owner
Retest affected scenarios
- Original conversation case
- Nearby variation and boundary
- Human handoff behavior
- Result in the destination system
Manage guides
Follow the sequence from queue operation and ownership through testing, context management and per-contact automation control.
Use the Conversation Inbox
Review messages, AI status, contact context and conversations that need human attention.
Read guide →Guide 02 · OwnershipAssign Conversations
Give work a clear person or queue and preserve a fallback for unassigned items.
Read guide →Guide 03 · VerificationTest Your AI Agent
Validate normal, ambiguous, difficult and escalation scenarios before release.
Read guide →Guide 04 · ContextUse Contact Fields and Notes
Store consistent facts and the human context needed for a reliable handoff.
Read guide →Guide 05 · OrganizationOrganize Contacts with Tags
Apply approved labels for filtering, workflow, source, status and coordination.
Read guide →Guide 06 · ControlPause the AI Agent for a Contact
Protect a human-led conversation without changing automation for every contact.
Read guide →Managing conversations
Which conversations should the team review first?
Start with human-attention requests, sensitive or uncertain cases and incomplete next actions. Then review assigned, important and newly active conversations before moving to quality patterns.
Does assigning a conversation mean it has been handled?
No. Assignment shows intended ownership. Confirm that the recipient reviewed the conversation and completed or advanced the approved next action.
When should I use a field, note or tag?
Use fields for consistent facts, notes for concise human context and tags for reusable operational labels. Store only information the workflow needs.
When should the AI Agent be paused?
Pause it for the contact when a person is actively taking over, the customer requests a person, or the case involves ambiguity, sensitivity, an exception or judgment outside the approved scope.
How should we correct a weak AI reply?
Preserve the example, identify the underlying knowledge or behavior source, change one approved source and retest the same scenario before releasing the update.
Send one conversation and the expected result
Include the ConvoRev account, channel, safe contact reference, conversation time, AI Agent, current owner or queue, observed behavior and expected next action. Remove credentials and unnecessary private customer information.