Create a call analysis response

AI-generated summary

Genesys Cloud can detect whether a live party or an answering machine answers the call. By default, the analysis begins when the call is answered, but you can configure it to begin on a beep using beep detection. When Genesys Cloud determines that an answering machine answered the call, it can hang up or transfer the call. You can also use adjustable live speaker detection (ADSL), which uses multiple methods simultaneously to determine quickly and efficiently whether the call is answered by a human. You can configure ADSL for higher speed or higher accuracy.

PrerequisitesClick to expand

Click Menu > Digital and Telephony > Outbound > Call Analysis Responses.
Click Create New.
Enter a unique and descriptive name in the Call Analysis Response Name box. This name can contain spaces and special characters, but must be unique.
If you do not want the system to enable call analysis after the call connects, select the Disable Post-Connect Call Analysis checkbox. When the call connects, the system disables all call analysis after the call connects and also disables Answering Machine Detection (AMD).
If you do not want the system to differentiate between a live voice and an answering machine, select the Disable Answering Machine Detection checkbox. When the call connects, the system transfers it without evaluating for a machine.
Note: If AMD is disabled, the system still accurately prevents around two thirds of voicemail and answering machines from going to the live voice response action. Several known default carrier messages can be identified almost immediately based on the voiceprint. Busy carrier announcements and tones, which typically account for about three percent of all calls, and faxes are also detected and prevented from taking the live voice response action.
If you selected a transfer option and want the system to delay the response until after a voicemail beep, select the Enable Beep Detection checkbox.
Notes:
- Enable Beep Detection is an option to enable when a voicemail or an answering machine is not being detected by AMD.
- Consider extending the no answer timeout on the campaign to 60 seconds due to the default ring settings on mobile phones to wait 60 seconds to detect the beep.
- In Architect, use the Detect Silence and Play Audio on Silence actions in your call flows to detect live or voicemail speech, and then determine how the flow proceeds.
If you want the system to transfer the call after a specific amount of time regardless of whether AMD is complete, select the Enable Answering Machine Timeout checkbox.
Under Start Answering Machine Detection Timer On, select when the timer starts for AMD to determine whether there is an answering machine or a live speaker on the line.
- Line Connect: The timer starts immediately when the line connects. This option processes calls quicker, but it allows for less accurate results.
- Speech Start: The timer starts when speech is detected. This option allows for more accurate results, but it processes calls slower.
For each of the call analysis results, Answering Machine and Live Voice, select one of the following actions for the system to take:
Note: For iPhone users from the United States, Canada, and Australia, who use American or Australian English Siri voices, call analysis can classify the Apple Call Screening prompt as live voice by default, and takes the associated action. To classify Apple Call Screening as an answering machine, contact Genesys Cloud Customer Care. The detection accuracy of Apple Call Screening decreases if you check the Disable Answering Machine Detection checkbox or if you set the adjustable live speaker detection level to anything other than disabled. For more information, see How does Genesys Cloud call analysis handle call screeners?. You can use tools available in Genesys Cloud to set up a specific flow for Apple Call Screening. For more information, see Differentiate call screeners from answering machines.
- Hangup disconnects the call.
- Transfer returns the call to the campaign’s queue so that the next available agent can answer it.
- Transfer to Outbound Flow transfers to the outbound flow you select. Create outbound flows beforehand using Architect.
  Note: Genesys does not recommend sending live voices to Architect flows, except for agentless campaigns. When you transfer live voices to a flow, the flow answers all calls immediately. This disrupts the pacing algorithm that manages call timing and agent availability. The dialer perceives all calls as answered and continues dialing faster to achieve the target abandon rate. However, since the flow answers all calls immediately, the target abandon rate is never achieved. The dialer continues to increase the dial rate until it reaches the calls-per-agent limit or another system limit, which causes excessive dialing and poor campaign performance. This also makes compliance abandon impossible.
Under Answering Machine Timeout, select one of the following actions:
- Hangup disconnects the call.
- Transfer returns the call to the campaign’s queue so that the next available agent can answer it.
- Transfer to Outbound Flow transfers to the outbound flow you select.
Under Answering Machine Timeout Duration, specify the duration of time AMD has to determine whether there is an answering machine on the line. The value is in milliseconds.

To select a different mode for improved call analysis performance, choose from the Adjustable Live Speaker Detection Level list.
To select a live speaker detection level, Post-Connect Call Analysis and Answering Machine Detection must be enabled. To enable Post-Connect Call Analysis, make sure the Disable Post-Connect Call Analysis checkbox is not selected. To enable Answering Machine Detection, make sure that the Disable Answering Machine Detection checkbox is not selected.

The speed of detection for a person most likely detected or a machine most likely detected can be selected for the following modes.

Each mode applies person-likely heuristics more aggressively, in terms of speed to live speaker detection, as it increases from Low to High. This also decreases accuracy in favor of speed to live speaker detection.

Mode	Description
Disabled	No change to current settings; standard timings are used.
Low	If Person-likely, then on post-connect generic speech detection (200–300 ms of speech), a live speaker is detected early. If the speech is a known audioprint with a strong match, then early live speaker detection is correctly avoided and the machine or busy is detected by the audioprint. (In this context, an audioprint refers to a pre-recorded carrier voicemail message or busy message.) If Machine-likely, then standard timings are used.
Medium	If Person-likely, then on line-connect, a live speaker is detected early. If Machine-likely, then standard timings are used.
High	If Person-likely, then on line-connect, a live speaker is detected early. If Machine-likely, then on post-connect generic speech detection (200–300 ms of speech), a live speaker is detected early. If the speech is a known audioprint (pre-recorded carrier voicemail message or busy message) with a strong match, then early live speaker detection is correctly avoided and the machine or busy is detected by the audioprint.

Notes:

As the mode increases, from Disabled through High, the live voice and Answering Machine Detection accuracy decreases, as there is less time for analysis, which increases the speed to detection.
Make sure that the external trunk setting Media Method is set to normal. For more information, see External trunk settings.

Adjustable live speaker detection (ALSD) also uses the number of ringbacks along with the audio to improve accuracy. Research suggests that if the number of ringbacks before an answer is either zero or higher than four, the answer is more likely to be a voicemail. If the number of ringbacks before an answer is between one and three, a live human connection is more likely.

ALSD mode	Person likely (1–3 rings)	Machine likely (0 or 4+ rings)
Disabled	Slow	Slow
1 (conservative)	Moderate	Slow
2 (moderate)	Fast	Slow
3 (aggressive)	Fast	Moderate

When ALSD detects a machine, it transitions to beep detection. The higher the ALSD mode, the faster ALSD can determine the response to be machine likely, and transition call analysis to beep detection, which optimizes beep detection time. If ALSD is disabled, call analysis only uses beep detection, which is slower, but more conservative.

After you assign an action to every call analysis result, click Save.
Note: An error message appears if the new call analysis response exceeds the maximum limit count of 1000. Delete old or obsolete call analysis responses as needed and click Save again to complete the new call analysis response configuration.

[NEXT] Was this article helpful?

Get user feedback about articles.

Mode

Description

Disabled

No change to current settings; standard timings are used.

Low

If Person-likely, then on post-connect generic speech detection (200–300 ms of speech), a live speaker is detected early. If the speech is a known audioprint with a strong match, then early live speaker detection is correctly avoided and the machine or busy is detected by the audioprint. (In this context, an audioprint refers to a pre-recorded carrier voicemail message or busy message.) If Machine-likely, then standard timings are used.

Medium

If Person-likely, then on line-connect, a live speaker is detected early. If Machine-likely, then standard timings are used.

High

If Person-likely, then on line-connect, a live speaker is detected early. If Machine-likely, then on post-connect generic speech detection (200–300 ms of speech), a live speaker is detected early. If the speech is a known audioprint (pre-recorded carrier voicemail message or busy message) with a strong match, then early live speaker detection is correctly avoided and the machine or busy is detected by the audioprint.

ALSD mode

Person likely (1–3 rings)

Machine likely (0 or 4+ rings)

Disabled

Slow

1 (conservative)

Moderate

Slow

2 (moderate)

Fast

Slow

3 (aggressive)

Fast

Moderate

Create a call analysis response

[NEXT] Was this article helpful?

Loading...

Create a call analysis response

[NEXT] Was this article helpful?