19.2. How to make use of AI rated calls

Natterbox AI allows you to automatically rate calls and perform various other analysis tasks on all of your business conversations. The next step is figuring out how to make the most of this valuable data.

This is one such way!

1.Viewing your overall business trends (Salesforce Dashboards)

Using the power of Salesforce Dashboards you’re able to take your Natterbox AI data and produce targeted charts and graphs that can provide great overview for the different areas of your business.

See here on how to create a report and dashboard of your choosing.

In this example we’re reporting on the average sentiment of the customers that our support team have been managing, this view quickly paints a picture of how content our customers are as well as highlighting our daily highs and lows.

This particular dashboard takes a birds eye view across all calls on the platform and takes an average of the score each day, but if we wanted we could build targeted dashboards and filter down on :

• Departments/Groups/Queues

• Specific Agents

• Specific Customer Accounts

• Specific campaign numbers

Also ‘Customer sentiment’ is just one example on the kind of data we can analyze, using the power of Natterbox AI it’s possible to craft prompts to produce a wide variety of data points such as (but certainly not limited to) :

• Agent performance

• Rating the different specific capabilities of the agent

  • Level of empathy

  • Level of professionalism

  • Questioning skills

Then beyond rating the call, it’s possible to pull out important information about a call such as:

• Was the issue solved on the call?

• Were our competitors discussed on the call?

• Was a cancellation discussed?

• Was a sale or booking made?

For more use cases see our library here.

2.Investigating anomalies (Reports)

Now you have your Dashboards you can quickly dig into any anomalies, such as the highlighted low points, from here you will be able to explore a Salesforce report of that particular area, a specific day for example.

From the report you’re able to easily see the outliers that have brought down the average that need investigating, from the above example you can see that there’s been a call with Customer Sentiment Rating of ‘1’.

Now we want to answer the question ‘why was this rating so low?’, now we could immediately hop onto step 3. and review the call, but in addition to that we can use the Salesforce report to provide us with more contextual data around this call to help identify the cause of it’s variance.

Here are a few usual fields that can be included within a report to help quickly provide background :

• Reasoning: The reason the AI provided the rating it did

  • This is the crux of the cause as the AI should be able to explain in human readable text as to why it gave it’s rating score

  • Note: this is an optional field, so you will need to ensure that it has been enabled on the Prompt itself, see more here

  • Also note within a Salesforce report the character limit is 255 characters, so in order to see the full reasoning you can either view it from the Insight Search page OR you can add the ‘AI Prompts Name’ field to the report to form a link to the AI Prompt record where the full reasoning can be viewed

• Answer Time: When was the call answered?

  • This can help identify whether the call was within busy peak hours for example.

• Call Summarization: A short summary of what happened on the call.

  • Great for providing a quick overview.

  • Note: again within a Salesforce report the character limit is 255 characters, so in order to see the full summary you can either view it from the Insight Search page OR you can add the ‘Insight Name’ field to the report to form a link to the Insight record where the full summary can be viewed.

• Natterbox User: Which agent handled the call?

• Account: Which Salesforce account was on the call?

  • Note: you can also look for Contact and Lead.

• Call Direction: Was the call was inbound or outbound? (whether the customer called in or was contacted by an agent)

• Category Details: Were there predefined Categories that triggered on the call? see more on them here

• Insight Player: A link to the full recording and transcription - see below for more.

From this example we can quickly gleam some key details:

1. The customer sentiment was given a rating of 1

2. The reasoning mentions that profanity was used by the agent

3. We see the summary mentions argumentation

4. We know the agent in question is Derek Smith

5. The affected customer is Travel Co

6. The Categories mention that this was a ‘Service Escalation’ and they backup the use of Profanity as the ‘Bad Language’ category was also triggered

This paints us a picture around what has happened and makes it clear this call does need to be investigated further, which brings us onto the next step.

3.Reviewing specific calls (Insight Player)

Click onto the Insight Player link to open the player for call anomaly.

From here you can listen to the full conversation and follow along with the transcription, from here you should be able to identify the reason behind the low rating.

4.Providing feedback to your agents (Notes)

Once you’ve highlighted the reason for the low call rating, you can then make a comment directly on the transcribed call and have it forwarded to the agent who dealt with the call for coaching support.

Mark timestamp

On the timeline click and pause the player header on the pivotal part of the conversation to obtain the timestamp

Click Add New Note

Categorize and Assign the Note

• Event Time: The time on the call in which the note is in context, this will use the timestamp of the play head by default, but it can be amended

• Relating to: You can leave a general comment (usually when there are no actions to be taken) or you choose to assign it to the agent directly so they can get notified

• Type of event: You can categorize the reason behind making the note

• Polarity: You can also further categorize the note by marking whether the note is positive/neutral/negative

See more on notes here, including creating your own event types and polarities.

Create Note

Write up the note you’re making

Add Note

Once complete you can click ‘Add Note’

Once created, the note will appear connected to the call and the note itself will be sent as a notification directly to the agent for review.