Real-time Agent Coaching

Real-time Agent Coaching helps supervisors guide agents during live calls. Agent AI analyzes utterances and speech patterns and triggers on-screen hints, nudges, or manager alerts when the configured conditions match. Supervisors can track triggered rules, measure adherence, and review patterns such as cross-talk, dead air, and speaking speed.

Rules

A rule combines a trigger (the condition) and an action (what happens when the condition matches). Each rule is independent, and you can turn it on or off, edit, delete, or tag it. Rule types:

Type	Description
Pre-built	Included with Agent Coaching. Two pre-built rules: None Intent and Price Objection.
Custom	Rules you create based on your contact center’s needs.

Pre-built Rules

None Intent

Prevents false positives during intent identification and trains the system to distinguish valid intents from ambiguous or incorrect ones. This rule is on by default. You can’t turn it off or delete it, but you can add utterances to improve accuracy. To add None Intent utterances:

Go to Agent AI > Agent Experience > Agent Coaching.
Select Edit on the None Intent pre-built rule.
Select Add Utterance(s).
Choose an input method:

Option A — AI Assist:

To use AI Assist, turn on OpenAI in Generative AI Tools and turn on Generating Opposite… under the Co-Pilots section.

Select AI Assist.
Select rules from the Add Rules list and select Add.
The AI generates similar utterances. Select the ones you want.
Select Add Selected, then Save.

Option B — Input Manually:

Select Input Utterances Manually.
Enter an utterance in the Utterances field and press Enter.
Repeat for additional utterances, then select Save.

Price Objection

A pre-built rule for handling customer price objections. You can customize the conditions and actions and turn it off, but can’t rename or delete it.

Setting	Value
Condition	Price objection utterance by customer
Alert type	Hint Agent
Alert title	Price Objection identified
Alert body	Please reassure the customer and offer discounts
Auto-close	No
Adherence	”Ok, Got it”
Assigned to	All Agents

Managing Rules

Create a Rule

Go to Agent AI > Agent Experience > Agent Coaching.
Select + New Rule.

Fill in the rule details:

Field	Description
Name	A clear name for the rule.
Description	Keywords and conditions that train the AI. Hover over the i icon for guidance.
Tags	Labels for grouping similar rules. One rule can have multiple tags.
Channels	Voice, Live Chat, or both.

Select Apply.
Select + Add a trigger to define when the rule triggers. See Triggers.
Select + Add an action to define what happens. See Actions.
Select + Add agents, groups to assign the rule. See Assign Rules.
Select Save.

Edit a Rule

Go to Agent AI > Agent Experience > Agent Coaching.
Search for the rule by its name or browse the list.
Select Edit on the rule.

Rules lock during editing — only one user can edit a rule at a time. If you leave a rule idle for 15 minutes or leave the screen, the lock releases automatically.

Delete a Rule

Go to Agent AI > Agent Experience > Agent Coaching.
Select Delete on the rule you want to delete.
Confirm by selecting Yes.

Turn on or off a Rule

Go to Agent AI > Agent Experience > Agent Coaching and select the Enabled toggle on any rule to turn it on or off. A confirmation message appears.

Optimize utterance matching for coaching rules

The Analyze feature compares the utterances in your coaching rules and pairs similar utterances across different rules. When two rules share highly similar utterances, the system can trigger both at once, creating ambiguity when it selects the next best action (NBA). Keep your utterance training data unique across rules so each coaching rule triggers as intended.

A similarity score near 1.00 means the two rules share duplicate training utterances. No threshold change can separate them — edit the training data instead.

Analyze utterance matches

Go to Agent AI > Agent Coaching.
Select Analyse.
Choose Overall Analysis or Custom Utterance.

Overall analysis

Overall analysis compares training utterances across all your coaching rules and flags rules with similar or overlapping matches. When you select Overall Analysis, the Overall coaching analysis page opens. The page lists ambiguous utterance pairs and sorts them by similarity score, highest first.

Custom utterance

Custom utterance analysis lets you test up to 10 utterances against your coaching rules before you add them to a rule. Test custom utterances

Select Custom Utterance.
Enter an utterance in the Utterances box.
Select + Add Utterance to add another utterance.
Select the X icon next to an utterance to remove it.
Select Analyse.

The results panel displays three tabs: Thresholds, Threshold summary, and Ambiguous pairs, and each tab header shows a record count. The Thresholds count equals the number of utterances you test.

Analysis results reference

Match fields — Thresholds and Ambiguous pairs tabs

Field	Description
Rule Name	Name of the coaching rule the utterance matched.
Trigger ID	Unique ID for the matched utterance’s trigger (prefix `aat-`).
Rule ID	Unique ID for the coaching rule (prefix `acr-`).
Utterance ID	Unique ID for the utterance.
Utterance	Text of the matched utterance.
Score	Similarity score between the utterances, from 0 to 1.
Threshold	One of the fixed threshold values the system tests for this match.

Summary fields — Threshold summary tab

Field	Description
Threshold	One of 11 fixed threshold values the system tests, from 0 to 0.9.
Top one correct	Percentage of utterances whose top match was the expected rule. Shows a dash if you didn’t set an expected rule.
Expected in Set	Percentage of utterances where the expected rule appeared anywhere in the results. Shows a dash if you didn’t set an expected rule.
No match	Count of utterances with no matching rule at this threshold.
Avg. rules	Average number of matching rules the system returns per utterance at this threshold.

The Custom utterance analysis dialog doesn’t support entering an expected rule ID. As a result, Top one correct and Expected in Set always show a dash for custom utterance tests.

Select the copy icon next to a value to copy that value. Select the copy response icon on the top-right of the page to copy the complete analysis response.

Optimize utterance matching for coaching rules

A similarity score near 1.00 means the two rules share duplicate training utterances. No threshold change can separate them — edit the training data instead.

Analyze utterance matches

Go to Agent AI > Agent Coaching.
Select Analyse.
Choose Overall Analysis or Custom Utterance.

Overall analysis

Custom utterance

Custom utterance analysis lets you test up to 10 utterances against your coaching rules before you add them to a rule. Test custom utterances

Select Custom Utterance.
Enter an utterance in the Utterances box.
Select + Add Utterance to add another utterance.
Select the X icon next to an utterance to remove it.
Select Analyse.

The results panel displays three tabs: Thresholds, Threshold summary, and Ambiguous pairs, and each tab header shows a record count. The thresholds count equals the number of utterances you test.

Analysis results reference

Match fields — Thresholds and Ambiguous pairs tabs

Field	Description
Rule Name	Name of the coaching rule the utterance matched.
Trigger ID	Unique ID for the matched utterance’s trigger (prefix `aat-`).
Rule ID	Unique ID for the coaching rule (prefix `acr-`).
Utterance ID	Unique ID for the utterance.
Utterance	Text of the matched utterance.
Score	Similarity score between the utterances, from 0 to 1.
Threshold	One of the fixed threshold values the system tests for this match.

Summary fields — Threshold summary tab

Field	Description
Threshold	One of 11 fixed threshold values the system tests, from 0 to 0.9.
Top one correct	Percentage of utterances whose top match was the expected rule. Shows a dash if you didn’t set an expected rule.
Expected in Set	Percentage of utterances where the expected rule appeared anywhere in the results. Shows a dash if you didn’t set an expected rule.
No match	Count of utterances with no matching rule at this threshold.
Avg. rules	Average number of matching rules the system returns per utterance at this threshold.

The Custom utterance analysis dialog doesn’t currently support entering an expected rule ID. As a result, Top one correct and Expected in Set always show a dash for custom utterance tests.

Select the copy icon next to a value to copy that value. Select the copy response icon on the top-right of the page to copy the complete analysis response.

Triggers

Triggers define when a rule activates. Every rule requires at least one trigger. Three trigger types are available.

Utterance Triggers

Triggers when a customer or agent says something matching the configured utterances. To set up an Utterance Trigger:

In the rule, select + Add a trigger > Utterance.
Select {select person} and choose Agent or Customer.
Select {set utterances} and enter an utterance. The AI suggests similar utterances with frequency percentages.
Select the utterances to include, then select Add.
Select Okay, then Save.
Set the frequency (how many times the utterance must occur).
Set the time frame:
Option Description
Anytime in conversation The utterance can occur at any point.
By time Within or from a specific number of seconds.
By number of messages Within or from a specific message count.
Press Enter to save.

Option	Description
Anytime in conversation	The utterance can occur at any point.
By time	Within or from a specific number of seconds.
By number of messages	Within or from a specific message count.

Speech Analysis Triggers

Triggers based on speech patterns such as cross talk, dead air, speaking speed, or talk ratio. To set up a Speech Analysis Trigger:

Select + Add a trigger > Speech Analysis.
Select {speech type} and choose a type.

Speech types:

Type	Description	Configuration
Cross Talk	Both parties speak simultaneously.	Set duration (seconds), frequency, and time frame.
Dead Air	Silence from the agent or customer.	Set duration, person (agent or customer), frequency, and time frame.
Speech Speed	Speaking rate in words per minute.	Set person, words per minute (default: 180), frequency, and time frame.
Talk Ratio	Agent-to-customer speaking time proportion.	Set comparator (Less than, Greater than, In between), percentage value, and time frame.

Cross Talk configuration:

Set duration, frequency, and time frame:

Dead Air configuration:

Speech Speed configuration:

Talk Ratio configuration:

Select a comparator:

Enter the percentage value:

A validation error appears if the configured duration is shorter than the selected conversation time period or if the trigger setup is incomplete.

Gen AI Utterances

Uses AI-powered prompts to analyze real-time conversations. When an utterance matches the configured prompt, the rule triggers. To set up a Gen AI Utterance trigger:

Select + Add a trigger > Gen AI Utterances.
Select {select person} and choose Agent or Customer.
Select {set prompt} and enter a prompt in the Enter Prompt box.
Select the tick mark to save.
Set the frequency and time frame.

Combining Triggers (And/Or)

You can combine multiple triggers on a single rule using the And or Or logic.

Select + Add a trigger to add a second trigger.
Choose And or Or.

Logic	Behavior
And	The rule activates only when all conditions match.
Or	Any one condition (or combined group) can trigger the rule.

Actions

Actions define what happens when a trigger activates. Three action types are available: Nudge Agent, Hint Agent, and Email Manager. Nudge and Hint actions support adherence tracking.

Adherence

Adherence tracks whether an agent follows through on a coaching instruction. You can set it up for Nudge and Hint actions.

Type	Description
Acknowledge Pressed	The agent acknowledges the message.
Utterance	The agent must say a specific utterance within a set time frame to confirm adherence.

Nudge Agent

Short, disappearing messages that appear on the agent’s screen and may or may not auto-close, depending on configuration. These messages compliment or redirect agents mid-conversation. Nudge types: Positive, Neutral, Alert, and Critical. To set up a Nudge Agent action:

Select + Add an action > Nudge Agent.
Select {Choose} and pick a nudge type (Positive, Neutral, Alert, Critical).
Select {Choose} and enter the nudge message. Optionally, select Use Gen AI Output to generate the message from a prompt, and add a fallback message.
Select the tick mark to save.
Set the display option:
Option Behavior
Auto-closes Closes after the configured duration (default: 5 seconds).
Doesn’t auto-close Stays on screen until the agent dismisses it.
Optionally, select + Add Adherence and choose Acknowledge Pressed or Utterance.

Option	Behavior
Auto-closes	Closes after the configured duration (default: 5 seconds).
Doesn’t auto-close	Stays on screen until the agent dismisses it.

Hint Agent

Pop-up messages that advise agents to take action. May or may not auto-close, depending on configuration. Hint types: Positive, Neutral, Warning, and Critical. To set up a Hint Agent action:

Select + Add an action > Hint Agent.
Select {Choose} and pick a hint type.
Enter a title. Optionally, use Gen AI Output to generate the title from a prompt, with a fallback (up to 42 characters).
Enter a message body. Optionally, use Gen AI Output with a fallback (up to 300 characters).
Set the display option:
Option Behavior
Auto Close Closes after the configured duration (default: 5 seconds).
Doesn’t auto close Stays on screen until the agent dismisses it.
Optionally, select + Add Adherence and choose Acknowledge Pressed or Utterance.

Option	Behavior
Auto Close	Closes after the configured duration (default: 5 seconds).
Doesn’t auto close	Stays on screen until the agent dismisses it.

Email Manager

Sends a push notification to a manager’s desktop for situations requiring supervisor awareness, such as repeated cross talk or customer profanity, instead of agent-facing nudges or hints. To set up an Email Manager action:

Select + Add an action > Email Manager.
Select {select people} and enter one or more manager email addresses (comma-separated).
Select the tick mark to save.
Select {set subject} and enter a notification subject.
Select {set body} and enter the notification content.
Select the tick mark to save.

Assign rules

By default, the system assigns rules to All Agents. You can narrow the assignment to specific agents or groups.

In the rule, select + Add agents, groups under Assign this to….
Select the agents or groups.
Select Save.

The rule appears under the group you created it for.

Rule Field Reference

Field	Description
Rule Name	Name of the rule and the users you assigned it to.
Actions	Edit or delete the rule.
Description	Short description of the rule’s purpose.
Tags	Labels you tagged the rule with.
Channels	Voice or Chat (or both).
Last Updated	Date and time you last modified the rule.
Status	Enabled or Disabled.

Speech Type Reference

Type	Description
Cross Talk	Both the agent and customer speak at the same time.
Dead Air	Silence from either party. Configurable by time frame: Anytime, First N seconds, or Within N seconds.
Speech Speed	Words per minute spoken by the agent or customer.
Talk Ratio	Proportion of speaking time by the agent relative to the customer. Configurable by percentage range and time period.

​Rules

​Pre-built Rules

​None Intent

​Price Objection

​Managing Rules

​Create a Rule

​Edit a Rule

​Delete a Rule

​Turn on or off a Rule

​Optimize utterance matching for coaching rules

​Analyze utterance matches

​Overall analysis

​Custom utterance

​Analysis results reference

​Optimize utterance matching for coaching rules

​Analyze utterance matches

​Overall analysis

​Custom utterance

​Analysis results reference

​Triggers

​Utterance Triggers

​Speech Analysis Triggers

​Gen AI Utterances

​Combining Triggers (And/Or)

​Actions

​Adherence

​Nudge Agent

​Hint Agent

​Email Manager

​Assign rules

​Rule Field Reference

​Speech Type Reference

Rules

Pre-built Rules

None Intent

Price Objection

Managing Rules

Create a Rule

Edit a Rule

Delete a Rule

Turn on or off a Rule

Optimize utterance matching for coaching rules

Analyze utterance matches

Overall analysis

Custom utterance

Analysis results reference

Optimize utterance matching for coaching rules

Analyze utterance matches

Overall analysis

Custom utterance

Analysis results reference

Triggers

Utterance Triggers

Speech Analysis Triggers

Gen AI Utterances

Combining Triggers (And/Or)

Actions

Adherence

Nudge Agent

Hint Agent

Email Manager

Assign rules

Rule Field Reference

Speech Type Reference