AI Call Scoring
A satisfaction score for every call, not just the ones that get a survey
EaseDial reads each finished agent call and predicts how satisfied the customer was, from 1 to 5, with the reasons. Supervisors see which calls and agents need attention without waiting for survey replies.
Definition
What AI call scoring is
AI call scoring is automatic quality review: an AI model reads the transcript of a finished call and scores it against a fixed rubric. In EaseDial, the result is a predicted CSAT score from 1 to 5 for each agent call, with a short explanation of why.
Traditional CSAT depends on customers answering a survey. Response rates are usually low, and the customers who do answer tend to be the very happy or very unhappy ones. Manual QA has the opposite problem: the scores are thorough, but a supervisor can only listen to a small sample of calls.
Predicted CSAT fills the gap. It covers every scored call, so you can see trends by agent and pick the right calls to review, then use human QA where judgement matters.
The Rubric
The five things every call is scored on
The same rubric is used for every call, which makes scores comparable across agents, queues and weeks.
Resolution
Did the customer get what they called for, or did the call end with the problem still open?
Courtesy and empathy
How the agent spoke to the customer: tone, acknowledgement of the problem, and patience.
Hold and wait handling
Whether holds were explained, kept short and followed up, rather than leaving the caller in silence.
Sentiment trajectory
Whether the customer sounded better or worse at the end of the call than at the start.
Escalation or complaint
Whether the customer asked for a manager, threatened to leave or made a formal complaint.
Workflow
How a call gets scored
Call ends
An agent call finishes and its transcript is saved to the call record.
Job picks it up
Every few minutes, a background job collects ended calls with transcripts that have not been scored yet.
Personal data redacted
Names, numbers and other personal details are redacted before anything is sent to the AI model. In HIPAA mode, nothing is sent.
AI scores the call
The model scores the call 1 to 5 against the five-factor rubric and writes down the reasons.
Score saved to the call
The predicted CSAT badge appears in call logs and the call journey, and rolls up into the agent report.
Reporting
Where scores show up
Each scored call shows a predicted CSAT badge in the call logs and the call journey. Scores roll up into the per-agent CSAT report under Reports, which includes:
CSAT % per agent
The share of an agent's scored calls rated 4 or 5, the same way survey CSAT is usually calculated.
Score distribution
How many calls landed on each score from 1 to 5, so one bad day doesn't hide behind an average.
Rubric averages
Average result for each of the five factors, which shows whether a dip comes from resolution, empathy or hold handling.
Coverage
How many of the agent's calls were actually scored, so you know how much weight to put on the numbers.
Privacy & Control
Built for teams that handle sensitive calls
Personal data is redacted from transcripts before they reach the AI model. For healthcare and other regulated teams, HIPAA mode keeps transcripts away from the AI completely.
Scoring is off until an admin turns it on, and every option below is set per company.
Admin controls
- ✓Enable or disable per company (off by default)
- ✓HIPAA mode: transcripts are never sent to the AI
- ✓Include or exclude internal calls
- ✓Let agents see their own scores
- ✓Daily cap on scored calls
In Practice
How to use predicted CSAT well
Use it to choose which calls to review. Sort an agent's calls by score and listen to the 1s and 2s first. The rubric reasons tell you what to listen for before you press play.
Coach on the factor, not the number. An agent whose average is pulled down by hold handling needs a different conversation from one whose resolution score is low.
Check coverage before comparing agents. A CSAT % built on 12 scored calls is far less reliable than one built on 300. The coverage figure in the report tells you which is which.
Calibrate against human QA. Have supervisors score a sample of the same calls with your QA scorecard. Where the two disagree, you learn something about the rubric or about your scorecard.
FAQs
AI Call Scoring FAQs
What is AI call scoring?
AI call scoring uses an AI model to review the transcript of a finished call and score it against a fixed rubric. In EaseDial the output is a predicted customer satisfaction (CSAT) score from 1 to 5 for each agent call, with the reasons behind the score, so you get a satisfaction signal on calls where the customer never answered a survey.
What is predicted CSAT?
Predicted CSAT is an estimate of how satisfied the customer was, made from what happened on the call instead of from a survey response. EaseDial scores resolution, courtesy and empathy, hold and wait handling, how the customer's sentiment changed during the call, and whether there was an escalation or complaint.
Is predicted CSAT as accurate as a real survey?
It measures something slightly different. A survey is the customer's own opinion but usually covers a small share of calls, and unhappy customers often skip it. Predicted CSAT covers every scored call but is an estimate. Most teams use it to spot patterns and pick calls to review, and treat individual scores as a starting point for a human listen, not a verdict.
Which calls get scored?
Ended agent calls that have a transcript. A background job picks them up every few minutes. Internal calls between colleagues are excluded unless you turn that option on, and you can set a daily cap on how many calls are scored.
What happens to customer data during scoring?
Personal data is redacted from the transcript before it is sent to the AI model for scoring. For healthcare and other sensitive teams there is a HIPAA mode that never sends transcripts to the AI at all.
Can agents see their own scores?
Only if you allow it. There is a setting that lets agents see their own predicted CSAT. Supervisors and admins see scores in call logs, the call journey and the per-agent CSAT report.
Is AI call scoring on by default?
No. It is off by default and is enabled per company by an admin. It is part of the AI quality features on EaseDial contact center plans; ask the sales team which plan fits your team.
How is this different from sentiment analysis?
Sentiment analysis labels how the customer felt (positive, neutral or negative), live and after the call. Predicted CSAT goes further: it scores the whole call on five factors, including whether the issue was resolved and how holds were handled, and explains the score. Sentiment trajectory is one of those five factors.
Related features
Related guides
See predicted CSAT on your own calls
Book a demo and we'll show the scoring rubric, the per-agent CSAT report and the HIPAA-mode settings.