Cisco UCCE Agent Greeting: Physical Phone vs. Mobile Agent Recording

[dongzhao2024]

We have a customer using Cisco NBR recording, and we’ve successfully identified the participant roles (visitor or agent) from both near-end and far-end audio—this works for both physical phones and mobile agents. However, we’ve noticed that the agent greeting segment is being detected incorrectly.

While digging into the issue, I asked ChatGPT to do some deeper research, and it provided a convincing explanation. I’ll be doing some verification on my end, but I’m also sharing it here for the community to review and comment.

================

Cisco UCCE Agent Greeting: Physical Phone vs. Mobile Agent Recording Behavior

Agent Greeting Overview in UCCE

Cisco Unified CCE’s Agent Greeting feature lets agents pre-record a greeting that plays automatically to callers on call connection​cisco.com. Under the hood, UCCE invokes an additional call leg to play this greeting via an IVR (e.g. CVP media server) when the agent answers. For local (physical) IP phone agents, the greeting is delivered using the phone’s Built-In-Bridge (BIB) as a conferencing mechanism. For Mobile Agents using Remote Call Control (RCP), which lack a local BIB, Cisco instead uses a network-based conference bridge to inject the greeting​cisco.com. This difference in implementation affects whether the greeting audio travels in the “near-end” (agent side) or “far-end” (caller side) audio stream, which is crucial for Network-Based Recording (NBR) systems like NICE.

Physical IP Phone Agents – Near-End vs. Far-End Audio

For an agent on a Cisco IP phone with BIB, the Agent Greeting is played through the phone’s bridge as if the agent’s device is originating the greeting media. When an agent answers, UCCE signals Unified CM to have the phone’s BIB place a call to the CVP media server and fetch the greeting prompt​cisco.com​cisco.com. The phone then mixes the greeting into the active call between the agent and customer​cisco.com. In practice, the greeting WAV file’s audio is sent from the agent phone to the caller (the phone transmits the greeting on the call leg to the customer) and is also audible to the agent in real-time​cisco.com.

From a recording perspective, this means the greeting is injected on the agent’s side of the audio path. The phone’s BIB effectively treats the greeting like agent-generated audio into the call. In a dual-stream recording (NBR) scenario, the near-end stream (agent->caller) will carry the greeting audio, since the phone is sending that audio to the caller​cisco.com. The far-end stream (caller->agent) during this time contains the greeting as heard by the agent, but that is the same media coming from the CVP leg into the phone. A properly configured recorder that forks both streams will capture the greeting on the agent’s channel. The key point is that the greeting becomes part of the agent’s outbound audio toward the customer. Cisco documentation confirms that for IP phone agents, the BIB is used to originate and mix in the greeting prompt​cisco.com​cisco.com.

Implications: In recordings of physical phone calls, the agent greeting segment will be present on the near-end (agent) audio track. If a recording solution only captured the far-end (customer) side, the greeting might be missed or only faintly heard (since the customer typically is not speaking during the greeting). Best practice is to record both streams and/or a mixed stream to ensure the greeting is captured. (In Cisco’s built-in recording API, both near and far audio are forked, so the greeting should be recorded as part of the agent’s transmitted audio.) Note that supervisors performing Silent Monitoring via BIB will not hear the greeting – this is by design​cisco.com​cisco.com, but it does not prevent the recorder from capturing it. The greeting is considered part of talk-time in reporting and is seamlessly recorded like agent speech when using phone-based forking​cisco.com​cisco.com.

Mobile (RCP) Agents – Near-End vs. Far-End Audio

For Mobile Agents (agents on a remote phone, controlled via RCP), the IP phone BIB is not available to inject the greeting. Instead, Unified CCE leverages an external conference bridge (usually a hardware DSP conference on a gateway) to play the greeting to the caller​cisco.com​cisco.com. In this call flow, when a mobile agent is connected, Unified CM will allocate a conference resource and an IVR port (CVP/VVB) into the call to deliver the greeting. Essentially, the call briefly becomes a three-party conference: the agent’s remote leg, the customer leg, and the greeting media (IVR) leg. The CVP media server plays the greeting into the conference, which both the caller and agent can hear (the agent may hear the tail end as a cue, though the call center may instruct agents to wait until it finishes)​cisco.com.

From the agent’s perspective, the greeting is coming from the network side. In other words, the greeting audio is injected into the conference by the IVR/bridge, not generated by the agent’s phone. Therefore, in an NBR setup the greeting will appear on the far-end (customer-side) audio stream of the agent’s call. The agent’s “near” side (their PSTN connection) is not the source of the greeting; instead, the media is flowing from the conference bridge towards both parties. A recording system must capture the customer-side media path to get this greeting segment. In practical terms, a mobile agent call often involves voice gateways for both the inbound call and the agent’s outbound leg. The greeting prompt might be injected on the CVP/gateway side and sent to the caller; the agent’s remote phone receives it via the CTI port bridge. Thus, a recorder tapping the network audio (far-end relative to the agent) will capture the greeting. If the recorder only looked at the agent’s outgoing PSTN audio, it would miss the greeting content, since the agent’s side is not where the prompt is originating in this case.

Cisco notes that enabling Agent Greeting for mobile agents consumes additional conference resources and even media termination points (MTP) if needed​cisco.com​cisco.com. This indicates the greeting is played via the Unified CM/gateway rather than the agent device. Additionally, if a Mobile Agent disconnects while the greeting is playing, the caller still hears the full greeting before the call terminates​cisco.com – further evidence that the greeting is coming from the network side (and is independent of the agent’s phone once initiated). In summary, for mobile/RCP agents the greeting audio lives on the far-end stream (network-injected media). The recording solution must be able to record that far-end audio; otherwise the greeting segment could be absent from call recordings.

Implications for Network-Based Recording (NICE and Cisco NBR)

Consistent Capture of Greeting Audio: Given the above, it’s critical that any NBR system (such as NICE Engage or Cisco’s built-in recorder) capture both the agent-side and customer-side RTP streams. In the local IP phone scenario, the agent’s stream carries the greeting, whereas in the mobile agent scenario the customer-side stream carries it. A properly configured system will fork or receive both streams and thus record the greeting in either case. If only a single side is recorded, one of the scenarios will lose the greeting audio. For example, a passive recording of just the agent’s PSTN leg would not include the greeting for mobile calls, and conversely a recorder focusing only on the network trunk might miss an agent phone’s BIB-injected greeting if it didn’t capture the agent’s transmit. Therefore, dual-stream recording is recommended, and the recording server should be aware of the call type to grab the correct media fork.

Cisco’s Network-Based Recording (NBR) feature (introduced in CUCM 10.x) was designed to handle such cases. NBR can use a CUBE or gateway to fork RTP for calls that don’t involve a local BIB, like off-network mobile agent calls​cisco.com​cisco.com. It will “automatically select the right media source based on call flow”​cisco.com – for a mobile agent call, this means forking at the gateway/trunk where the conference media is present. In a NICE recording deployment, this typically means using Cisco’s recording integrations (JTAPI or built-in-bridge control) combined with gateway forking for remote calls. To ensure the Agent Greeting is captured for mobile agents, the recording solution should leverage this network recording on the trunk or conference resource. Cisco documentation confirms that with network-based recording, calls extended off-network (e.g. to home or mobile agents) can still be recorded regardless of device​cisco.com.

Audio Stream Placement: It’s worth noting that in both scenarios the greeting is audible to both parties, but its point of injection differs. For a physical phone agent, the recorder will see the greeting on the agent’s channel (since the IP phone sent that audio to the caller)​cisco.com. For a mobile agent, the recorder will see the greeting on what corresponds to the caller’s channel (since the media came from the CVP/bridge towards the agent and caller)​cisco.com. As a result, quality monitoring or speech analytics systems should be configured with this in mind – e.g. if analyzing the agent’s speech segment, the greeting (though spoken by the agent in the recording) is actually a fixed pre-recorded file. Some organizations choose to tag or segregate the greeting audio in recordings, since it’s not the agent’s live speech. Knowing which channel to find the greeting on for each call type can assist in such post-processing.

Configuration and Best Practices for Capturing Agent Greetings

To achieve consistent recording of Agent Greeting audio, consider the following configuration details and best practices:

• Enable Built-In Bridge on Agent Phones: For local agents, ensure the IP phone’s Built-In Bridge is enabled in CUCM (set to “On” or “Default (On)” per device)​cisco.com. The Agent Greeting feature requires BIB on supported phones; if BIB is disabled, CUCM will attempt to fall back to a conference bridge for greeting injection​cisco.com. That can alter which audio stream the greeting uses and potentially complicate recording. Always use supported phone models/firmware that have BIB, as listed in the compatibility matrix​cisco.com.

• Configure Conference Resources for Mobile Agents: In a mobile agent deployment with Agent Greeting, you must configure hardware conference bridge resources in CUCM and the gateway (DSP farm)​cisco.com. This allows CUCM to inject the greeting via a conference for remote agents. Insufficient conference resources can lead to greeting failure or no audio. Cisco recommends sizing the number of conference ports based on call rate and average greeting length​cisco.com (since each greeting uses a port for a short time). Also, if the mobile agent’s call comes in over a SIP trunk from another cluster, ensure an MTP is available or use SIP endpoints, so the conference can be established without media issues​cisco.com.

• Recording Profile and Media Forking: When using CUCM’s built-in recording (for example, with NICE), apply a Recording Profile that forks both near-end and far-end streams to the recorder. In CUCM, this means two SIP legs (one for each stream) will be sent to the recording server​cisco.com. This dual-fork is crucial – as discussed, the near-end fork contains the greeting in one case and the far-end fork contains it in the other. Verify with your recording vendor’s documentation (e.g. NICE Uptivity or Engage) that it supports capturing both channels for Cisco contact center calls. Most Cisco-certified recording solutions do this by default.

• Network-Based Recording for Remote Calls: Leverage network-based recording for scenarios where the agent is not on a Cisco phone. CUCM’s NBR can instruct a CUBE or gateway to fork audio for mobile agent calls​cisco.com​cisco.com. Make sure the CUBE is configured with the Extended Media Forking (XMF) API and added to CUCM as a recording-enabled gateway. This ensures that even if the agent is offsite, the greeting (and the rest of the call) is forked to your recorder from the network side. NBR can record calls “regardless of device, location, or geography” by using the gateway as the tap point​cisco.com – an ideal solution for mobile agent audio capture.

• Testing and Tuning: After enabling Agent Greeting, perform test calls with both local and mobile agents and check the recordings. Confirm that the greeting is present and clear in the recorded audio. If the greeting is low or missing on one channel, you may need to adjust which stream the recorder is using or confirm the forking configuration. Cisco best practice is to monitor the greeting playback levels and timing – e.g. ensure the agent is not speaking before the greeting ends. This has indirect effects on recording quality (to avoid agent voice overlapping the greeting).

In summary, for physical IP phone agents, the Agent Greeting is injected via the agent’s BIB and travels in the agent’s (near-end) stream to the customer (also audible in the agent’s earpiece)​cisco.com​cisco.com. For mobile agents, the greeting is delivered via a network conference bridge and thus travels in the far-end stream from the network to both caller and agent​cisco.com. A robust NBR solution must capture both directions to reliably record the greeting in all cases. By enabling the correct Cisco features (BIB, conferencing) and using the appropriate recording integration, you can ensure the Agent Greeting audio is consistently captured on call recordings. This guarantees that QA reviewers and compliance logs include the full customer experience, greeting included, for both local and remote agents.

Sources:

• Cisco UCCE Features Guide – Agent Greeting (explains BIB usage and call flow)​cisco.com​cisco.com

• Cisco Packaged CCE Guide – Mobile Agent (notes external conference requirements for greetings)​cisco.com​cisco.com

• Cisco CUCM/JTAPI Documentation (agent greeting audible to both parties; BIB initiates IVR leg)​cisco.com​cisco.com

• Cisco Support Document – Network-Based Recording for Mobile Agents (NBR records off-network calls via gateway)​cisco.com​cisco.com

• Cisco Solution Design Guide – UCCE 12.5 (BIB required; fallback to conference if disabled)​cisco.com and call flow description​cisco.com.

[bill.king1]

Hi, not sure what you're trying to accomplish/saying when you say that the agent greeting is being detected incorrectly?
If you use call recording using built in bridge for instance for traditional non-mobile agent, the agent greeting is recorded as part of the call as expected.

[dongzhao2024]

Thanks @bill.king1 for the quick response! We use NBR to record calls for both physical phones and mobile agents, with gateway preferred approach. The recording also include internal agent to agent calls which are captured by built-in bridge. Our problem is for those recorded call media forking, we need to distinguish which audio is visitor, which audio is agent based on the call flow and farend/nearend attribute, and we noticed unexpected behaviour for agent greeting part. The above info is trying to find a explanation for this behaviour and verify it accordingly. Not sure whether others have similar issue before.