Ai CSAT (Customer Satisfaction)

  • Updated

Let's dig into the details. 

How Ai CSAT Works

Ai CSAT uses the latest deep learning technology to detect Customer satisfaction from your Dialpad call transcripts between Agents and Customers.

First, the audio from the call is transcribed into words by our proprietary speech recognition technology to maximize accuracy in noisy environments.  Then, the transcript is fed into a deep neural network that makes accurate predictions by updating its ~100 million parameters based on the 1.3 million anonymized calls used for training.

Our algorithm learns as it collects more data, and can detect things like:

    • If the customer was put on hold for too long
    • Did the Agent continuously interrupt the customer
    • Was the customer's issue resolved or left unresolved
    • What was the Customer's sentiment from the beginning to the end of the interaction
    • If the customer uses positive language: 
      • “You’ve been so helpful, thank you”
      • “You solved it! You’re the best!”
    • If the customer uses negative language:
      • “I was on hold too long”
      • “I want to cancel”
      • “Your audio keeps cutting out”


The result is a highly precise model that can predict CSAT from the interactions in the call transcript — with an 87% accuracy rate!  

Purchase Ai CSAT

To add Ai CSAT to your plan, head to 

  1. Navigate to Preferred Add-ons 
  2. Select Purchase licenses beside Ai CSATbuy
  3. Choose if you'd like to apply BI to your Contact Center or Ai Sales plan (or both!)add_ai_csat_to_which_plan.png
  4. Review the total cost and Confirm Purchase

That's it,  you're all set!

Once Ai CSAT has been added, there's no additional setup needed, you just need to make sure Ai is enabled on your Contact Center.

Ai CSAT Analytics

CSAT analysis helps companies improve customer satisfaction, loyalty, retention, and revenue by identifying areas for improvement through surveys and feedback from customers. Get granular with your CSAT data in Dialpad's built-in Ai CSAT analytics dashboard.

To view your Ai CSAT data:

  1. Head to
  2. Select Ai CSAT from the left-hand sidebar OR select the Ai CSAT tab in the Analytics overview


Let's go over each section of the Ai CSAT dashboard.

CSAT Overview

Here you'll see your average CSAT over a 30-day period, and how it compares to the previous 30 days.

You'll also see your response rate for traditional CSAT surveys, as well as the percentage of Ai CSAT graded calls. 


CSAT Over Time

The CSAT Over Time section shows your daily (or weekly) breakdown of CSAT responses. 

Click on a day to view the daily average and number of graded calls. 


Agent Leaderboard

The Agent leaderboard displays each Agent's average CSAT score, their number of graded calls, and their percentage change (good or bad) as compared to the 30 days prior. 


Weekly Trend Emails

Dialpad makes it easy to stay on top of your performance. We'll email you a Weekly Ai CSAT Report to highlight trends such as your top performers, week-over-week changes, number of scores, and more.




What languages does Ai CSAT support?

Dialpad's Ai CSAT surveys are currently available only in English.

How accurate is Ai CSAT?

Dialpad Ai learns from your customer interactions and improves over time. It's currently sitting at a healthy 87% accuracy rate, but continues to improve every day!

If you are using CSAT surveys through IVR/DTMF or SMS, the accuracy will be the highest.

This acts as training data to help tailor the algorithm to your customer interactions.

That's why we recommend always requesting CSAT from your customers, even if response rates are low.

Why don’t all my calls have CSAT scores?

Ai CSAT calculates CSAT scores for calls that don’t have survey responses. All calls are evaluated by Ai, however not all calls merit a score.

In order for Ai to grade a call the following parameters must be met:

  • The Agent must have Contact Center or Ai Sales license
  • The call must be in English
  • The call cannot be a voicemail
  • the call must be inbound
  • The call was completed — if the Agent still needs to follow up (for example if they say "I'll call you back" or " let me give you a call back") the call is not processed
  • The sentiment is clear. Sometimes it's just tough to tell — if our Ai is not confident, we prefer to say it’s neutral

On export, why is Ai CSAT a "5" or a "1"?

As of August 1st, 2022, Ai CSAT predicts satisfaction on a scale of 1 to 5 based on how confident we are in the satisfaction score for the interaction. Exports after this date will include calls with 1,2,3,4 or 5 as the CSAT.

Prior to August 1st, 2022, Ai CSAT scored calls as satisfied, unsatisfied, or neutral. We represent these as 5, 1, or none respectively to make it easy for you to calculate your CSAT score in Analytics as the percent of satisfied customers over all rated calls.

Why can't I find Ai CSAT data prior to Sept 22, 2021 in Analytics?

Customers with Dialapad Ai enabled before Sept 22, 2021 will see predicted CSAT calls on and after this date. This is the first date we launched the feature.

Did I notice a change in average CSAT from Ai CSAT? What could it be?

To ensure continuity and accuracy, we do our best to “pin” the average CSAT predicted by Ai to your customer-collected averages for CSAT. For customers who do not collect customer satisfaction, there’s no information to ground the algorithm to so algorithm accuracy improvements can cause a change in average CSAT.

If you noticed a change in CSAT score on the following dates, this could be due to the release of a new more accurate algorithm:

    • Oct 21, 2021 - Accuracy improvement
    • Aug 1, 2022 - Ai CSAT now predicts satisfaction on a scale of 1 to 5 based on how confident it is in the satisfaction score for the interaction. More calls with a neutral score of 3 can bring down the CSAT score because they are not satisfied customers (see our FAQ: What is the CSAT score when it's shown as a percentage?)