Setting up last agent and preferred agent routing in Webex Contact Center

Do

Cisco's 2021 Setup and Administration Guide describes Agent-based Routing as an optional feature 19. Queue To Agent returns failure code 4, FEATURE_NOT_ENABLED, when the Agent-based Routing feature is not enabled 35 32. Decide which external system holds the contact-to-agent mapping, because for both the preferred-agent and last-agent use cases Cisco says the contact details and the mapping are stored outside Webex Contact Center 22. A contact can be mapped to one or more preferred agents 23.

Verify

Queue To Agent writes QueueToAgent.FailureCode as an output variable 38, and failure code 4 means the Agent-based Routing feature is not enabled 35. Suggested check: during the end-to-end test calls, confirm the stored failure code never shows the feature-disabled value.

Do

Pick an existing queue as the reporting queue: its configuration, such as recording and default music in queue, applies to the preferred-agent contact, and the contact is reported against it 46 41. Pick a recovery queue for contacts that Queue To Agent fails to deliver, that the agent does not answer, or that the agent rejects 40. Choose a recovery queue whose agents can take any of these callers, because in the recovery queue the contact goes to the longest available agent and skills-based routing is not supported 45. Pick the regular queue that a chained Queue Contact activity will use when no preferred agent is available 30.

Verify

Queue To Agent failure code 5, INVALID_VTEAM_ERROR, indicates an invalid reporting or recovery queue 32. Suggested check: when you configure the preferred-agent activity, confirm both chosen queues appear in its queue drop-downs.

The last agent who handled this caller

Do

In Control Hub go to Services > Contact Center > Tenant Settings > Integrations > Connectors, choose the Webex Contact Center card, give the connector a unique name, and tick the authorisation checkbox 56. Read-Only Access allows GET API calls 5, and Read-Write Access allows GET, POST, PUT, UPDATE and DELETE calls 6. WarmTransfer's reading of the sources is that the connector article does not say which access level the Search API needs, so a Search call sent as POST may need a Read-Write connector and should be tested before you rely on Read-Only 4.

Verify

In Cisco's developer blog example, the HTTP Request activity has Use Authenticated Endpoint turned on and the Webex Contact Center connector selected 2. Suggested check: open a lookup activity in the flow editor, turn on the authenticated endpoint option, and confirm the new connector appears in the connector list.

Rollback

Suggested rollback: remove the connector from every flow that uses it, then delete it.

An owner recorded in a CRM or another external system

Do

In Control Hub go to Services > Contact Center > Tenant Settings > Integrations > Connectors and create a custom connector whose Resource Domain URL starts with https:// 8. Use Basic Authentication, the default, or OAuth 2.0 with client credentials and a certificate, client credentials and a client secret, or password grant 7. An OAuth 2.0 custom connector requires Client ID, Client Secret, Token URL and Scope fields 62. Check how many custom connectors already exist, because an organisation can add up to 10 9.

Verify

The HTTP Request activity has Use Authenticated Endpoint and Connector fields 17. Suggested check: open a lookup activity in the flow editor, turn on the authenticated endpoint option, and confirm the new custom connector appears in the connector list.

Rollback

Suggested rollback: delete the custom connector once no flow references it. An organisation can add up to 10 custom connectors, so deleting an unused one leaves room under that limit 9.

The last agent who handled this caller

Do

Cisco says the preferred-agent lookup is done with the HTTP Request flow activity 21, and Cisco's developer blog says the Search API, called from Flow Designer, can look up the last agent a caller spoke with so the flow can queue the caller to that agent 50. Configure the HTTP Request as in Cisco's blog example: Use Authenticated Endpoint on, the Webex Contact Center connector selected, and /search as the Request Path 2. Write the body as a taskDetails GraphQL query, which runs over Customer Session Records and Customer Activity Records 16. Include the mandatory from and to arguments as Long epoch timestamps 13. Filter origin with the equals operator against the caller's number in E.164 form 15. Select lastAgent with its id subfield 14. When a query fetches AAR or CAR data, the span between from and to cannot exceed 30 days 12. WarmTransfer's reading of the sources is that the last-agent lookup should keep from-to within 30 days and sort or limit to the most recent task 51. Use Parse Settings, an Output Variable and a Path Expression, to put the agent ID into a flow variable 18. Flow Designer's templates include one called "Last agent routing template", which you can start from 52.

Verify

The origin filter compares against a phone number in E.164 form, such as +12344565678 15. Suggested check: in a test call from a number that has called before, confirm the parsed variable holds an agent ID, and confirm it is empty for a number that has never called. Suggested check: confirm the caller number in the flow uses the same format as the stored origin value.

Rollback

Suggested rollback: delete the lookup activity and republish the flow.

An owner recorded in a CRM or another external system

Do

Cisco says the preferred-agent lookup is done with the HTTP Request flow activity, which retrieves the mapping from an external application 21. Turn on Use Authenticated Endpoint, select the CRM connector, and set the Request Path, Method and Content Type for the CRM's lookup endpoint 17. Use Parse Settings, an Output Variable and a Path Expression, to put the owner into a flow variable 18. Store either the agent's email address or ID, because the Queue To Agent Agent Variable holds one of the two 28. Review the Response Timeout and Number of Retries fields on the activity 17.

Verify

Suggested check: in a test call for a known customer record, confirm the parsed variable holds the expected owner.

Rollback

Suggested rollback: delete the lookup activity and republish the flow.

The last agent who handled this caller

Do

Add Queue To Agent after the lookup and set Agent Variable to the flow variable filled in step 4 28. Agent Lookup Type is either Email or ID 36. We infer that a last-agent flow should set Agent Lookup Type to ID, because the Search API returns the last agent as lastAgent.id 20. Select the reporting queue by queue ID from its drop-down 41, and select the recovery queue chosen in step 2 40. Set a static priority from P1, the highest, to P9, the lowest, which ranks contacts queued to the same agent 42. Alternatively, use a flow variable that returns an integer from 1 to 9 44.

Verify

Queue To Agent writes QueueToAgent.AgentId and QueueToAgent.FailureCode 38, and failure code 2 is AGENT_NOT_FOUND 32. Suggested check: place a test call from a returning number and confirm the agent output holds the agent who last handled that number.

Rollback

Suggested rollback: remove the activity and reconnect the lookup output to your existing queueing activity.

An owner recorded in a CRM or another external system

Do

Add Queue To Agent after the lookup and set Agent Variable to the flow variable filled in step 4 28. Set Agent Lookup Type to Email or ID to match what the CRM returns 36. Select the reporting queue by queue ID from its drop-down 41, and select the recovery queue chosen in step 2 40. Set a static priority from P1, the highest, to P9, the lowest 42, or a flow variable that returns an integer from 1 to 9 44. If the CRM can return a second owner, chain a second Queue To Agent activity for that agent, because a flow can chain Queue To Agent activities to try consecutive preferred agents 29.

Verify

Queue To Agent writes QueueToAgent.AgentId and QueueToAgent.FailureCode 38, and failure code 2 is AGENT_NOT_FOUND 32. Suggested check: place a test call for a known customer and confirm the agent output holds the owner the external system names. Suggested check: if the agent is not found, compare the email address or ID stored in the external system with the agent's record.

Rollback

Suggested rollback: remove the activity or activities and reconnect the lookup output to your existing queueing activity.

Do not wait and go straight to the queue

Do

Leave the "Park Contact if Agent is unavailable" toggle off 39. With the toggle off, a contact whose preferred agent is unavailable fails to reach the agent 37, and Queue To Agent takes its failure path 33.

Verify

Failure code 1 is AGENT_UNAVAILABLE, and failure code 6, AGENT_BUSY, means the agent is available but already engaged in another call 32 34. Suggested check: with the preferred agent not available, place a test call and confirm the contact leaves the preferred-agent activity by its failure path.

Wait a fixed number of seconds and then go to the queue

Do

Turn the "Park Contact if Agent is unavailable" toggle on, which parks the contact against the preferred agent until that agent becomes available 39. WarmTransfer's reading of the sources is that a timed fallback can be built by parking the contact with Queue To Agent, playing music for a set Music Duration, and then running Queue Contact 53.

Verify

Suggested check: with the preferred agent not available, place a test call and confirm the contact is parked rather than failed, and that the activities after the preferred-agent activity run.

Rollback

Suggested rollback: turn the park toggle off and republish the flow.

Wait for the preferred agent however long it takes

Do

Turn the "Park Contact if Agent is unavailable" toggle on 39. If the preferred agent is unavailable and the park option is selected, the contact is parked and waits for that agent to become available 26. WarmTransfer's reading of the sources is that none of them gives a maximum park time, so parking with no flow-side timer may leave a caller waiting as long as the preferred agent stays unavailable 25.

Verify

Suggested check: confirm the business has accepted an open-ended wait for these callers before going live. Suggested check: with the preferred agent not available, place a test call, confirm it stays parked, then make the agent available and confirm the call is offered to them.

Rollback

Suggested rollback: turn the park toggle off and republish the flow.

Do not wait and go straight to the queue

Do

Connect the Queue To Agent failure path to a Queue Contact activity that targets your regular queue 30 33. Connect the HTTP Request and Queue To Agent error-handling paths to the same Queue Contact, because an activity with no error handling path configured falls back to the path defined in the OnGlobalError event handler 63. The Undefined Error node is not present on the HTTP Request activity 64. Field reports suggest the activity set uses at least 5 label conventions for an outcome, so check each activity's own port names 65.

Verify

Suggested check: run the unknown-caller, logged-out-agent and unavailable-agent tests and confirm each ends in the regular queue.

Rollback

Suggested rollback: remove the added connections, but leave the failure path connected to a valid destination.

Wait a fixed number of seconds and then go to the queue

Do

After Queue To Agent, add a Play Music activity and set Music Duration to the wait you chose, in seconds 27. Then add a Queue Contact activity that targets your regular queue 30. WarmTransfer's reading of the sources is that Cisco documents this chaining and the duration setting but not how Queue Contact treats a contact that is still parked against an agent, so treat the design as unproven until the step 10 tests pass 53. Connect the failure path to the same Queue Contact, because it is still taken when the variable lookup cannot find the preferred agent 33. Connect the HTTP Request and Queue To Agent error-handling paths there too, because an activity with no error handling path configured falls back to the OnGlobalError event handler path 63. Field reports suggest the activity set uses at least 5 label conventions for an outcome, so check each activity's own port names 65.

Verify

Suggested check: with the preferred agent not available, confirm the caller hears music for the chosen wait and then reaches an agent in the regular queue. Suggested check: make the preferred agent available partway through the music and confirm that agent gets the call.

Rollback

Suggested rollback: remove the music and fallback queue activities, then turn parking off as in the no-wait option.

Wait for the preferred agent however long it takes

Do

The reporting queue's configuration, such as default music in queue, applies to the preferred-agent contact, so set the parked caller's music on that queue 46. Connect the Queue To Agent failure path, which is taken when the variable lookup cannot find the preferred agent, to a Queue Contact activity that targets your regular queue 33 30. Connect the HTTP Request and Queue To Agent error-handling paths there too, because an activity with no error handling path configured falls back to the OnGlobalError event handler path 63. Contacts that the agent does not answer or rejects go to the recovery queue without extra wiring 40.

Verify

Suggested check: confirm an unknown caller lands in the regular queue, and a parked caller hears music until the preferred agent is free.

Rollback

Suggested rollback: remove the added connections and activities, then turn parking off.

Do

Review the RONA (Redirection on No Answer) timeout, which is set at tenant level in Control Hub under Services > Contact Center > Tenant settings > Desktop 49. For telephony it defaults to 18 seconds, with a minimum of 1 second and a maximum of 120 seconds 48. WarmTransfer's reading of the sources is that the RONA timeout is the most likely bound on how long a preferred agent rings before the contact moves to the recovery queue 47.

Verify

Suggested check: during the end-to-end tests, time an unanswered offer to the preferred agent and compare it with the configured timeout.

Rollback

Suggested rollback: restore the previous timeout value.

Do

Store QueueToAgent.FailureCode, QueueToAgent.AgentState and QueueToAgent.AgentIdleCode where you can see them later 38. A Webex developer blog lists Global Variables and Caller Associated Data (CAD) Variables among the categories of flow data 66. Global variables are created by an administrator in the Provisioning module of Control Hub and then consumed inside a flow 67. If you need actions when the agent accepts or disconnects, handle the AgentAccepted and AgentDisconnected events that Queue To Agent triggers in the main flow's Event Flows tab 31. Expect preferred-agent contacts to be reported against the reporting queue 41. When a Callback activity follows Queue To Agent, customers can request a callback from their preferred agent 68. The Analyzer Callback Report's Preferred Agent Name column shows N/A when the contact was not queued to a preferred agent through Queue To Agent 3.

Verify

Suggested check: after the end-to-end tests, confirm the stored failure codes match each scenario and the test contacts appear under the reporting queue.

Rollback

Suggested rollback: remove the variable assignments and republish the flow.

Do not wait and go straight to the queue

Do

Place these test calls and compare the outcome with the Queue To Agent failure codes and recovery behaviour 32 40:

• A caller whose preferred agent is available should connect to that agent 1.
• A caller whose preferred agent is unavailable should show failure code 1, AGENT_UNAVAILABLE, and then reach the regular queue 32 30.
• A caller whose preferred agent is logged out should show failure code 3, AGENT_NOT_LOGGED_IN, and then reach the regular queue 32.
• A caller whose preferred agent is on another call should show failure code 6, AGENT_BUSY 34.
• An unknown caller should show failure code 2, AGENT_NOT_FOUND, or an empty lookup, and then take the failure path to the regular queue 32 33.
• A preferred agent who lets the call ring out or rejects it should cause the contact to go to the recovery queue 40.

Verify

Suggested check: confirm every case ends with an agent and no caller hears a disconnect.

Wait a fixed number of seconds and then go to the queue

Do

Run the same cases as for the no-wait option, comparing outcomes with the Queue To Agent failure codes and recovery behaviour 32 40:

• A caller whose preferred agent is available should connect to that agent 1.
• A caller whose preferred agent is logged out should show failure code 3, AGENT_NOT_LOGGED_IN 32.
• A caller whose preferred agent is on another call should show failure code 6, AGENT_BUSY 34.
• An unknown caller should take the failure path to the regular queue 33 30.
• A preferred agent who lets the call ring out or rejects it should cause the contact to go to the recovery queue 40.
• A caller whose preferred agent stays unavailable for longer than the Music Duration should reach the regular queue, which tests the timed-fallback design 27 30.
• A caller whose preferred agent becomes available partway through the music should reach that agent, because a parked contact waits for that agent 39.

WarmTransfer's reading of the sources is that these timing cases are what prove or disprove the timed fallback, because Cisco does not document how Queue Contact treats a contact still parked against an agent 53. Also check that no test returns failure code 7, VTEAM_TRANSITION_LIMIT_REACHED 32.

Verify

Suggested check: confirm both timing cases behave as expected. Suggested check: if a parked contact is not moved to the regular queue after the music, switch to the no-wait option.

Wait for the preferred agent however long it takes

Do

Run the same cases as for the no-wait option, comparing outcomes with the Queue To Agent failure codes and recovery behaviour 32 40:

• A caller whose preferred agent is available should connect to that agent 1.
• A caller whose preferred agent is unavailable should be parked and wait for that agent 26.
• A caller whose preferred agent is logged out should show failure code 3, AGENT_NOT_LOGGED_IN 32.
• An unknown caller should take the failure path to the regular queue 33 30.
• A preferred agent who lets the call ring out or rejects it should cause the contact to go to the recovery queue 40.

Add a long-wait case in which the preferred agent stays unavailable for several minutes, and note whether the park ends on its own; WarmTransfer's reading of the sources is that none of them gives a maximum park time 25.

Verify

Suggested check: confirm parked callers are offered to the preferred agent as soon as that agent is available. Suggested check: confirm the long-wait experience matches what the business accepted when choosing this option.

Do

Turn validation on and publish the flow, because the Publish flow button stays disabled while the Validation toggle is off or any active errors exist 69. Flow validation checks only structural errors and cannot confirm that variables resolve to expected values at runtime 70. Flow Designer has no version control feature: the version number counts publishes, and version labels can be applied 71. Attach a version label such as Live, Test or Dev when publishing; "Latest" is applied by default to the most recently published version 72. Keep a record of the previous design before you change the live flow, because Flow Designer has no version control to restore it 71.

Verify

Suggested check: confirm real calls on the entry point follow the new path, and repeat one of the end-to-end test cases on the live number.

Rollback

Suggested rollback: rebuild the previous design from your record, removing the lookup, the preferred-agent activity and the fallback wiring, then republish. Do not rely on undo, because Flow Designer undo and redo covers only the last 10 canvas actions 73.