What is a customer satisfaction score?

Customer satisfaction score (CSAT) is a survey measure that asks customers to rate how satisfied they were with an interaction, product or service, typically on a short numbered scale right after a conversation ends. It is the most common support metric. Applied to AI agents, it shows how the conversation felt, not whether the agent was right.

Why does CSAT matter once agents act?

Customers judge an agent the way they judge a person: was it quick, polite, clear and helpful? CSAT captures that. A falling score is an early warning that an agent is frustrating people, even when internal measures look fine.

The danger runs the other way. A customer who was told what they wanted to hear will rate the conversation highly, even if what they heard was wrong. An agent that promises a refund it cannot authorise, confirms a fee waiver that does not exist or tells a customer a claim is covered before it has been assessed can score very well. The error surfaces later, as a complaint.

How is CSAT measured in practice?

The firm sends a short survey after the contact, often a single question with an optional comment. Results are reported as an average, or as the share of respondents who chose the top ratings.

Three cautions apply. Response rates are low and skewed towards very happy and very unhappy customers. Scores reflect the outcome as much as the service, so a customer refused a refund may rate a well-run conversation poorly. And CSAT says nothing about the many customers who did not respond. For agents, report CSAT by intent and by whether the agent handled the case alone or handed over.

An example from a lender's support chat

Take a consumer lender whose chat agent scores well on CSAT for payment date changes. One week the score rises further. The team investigates and finds that after a source update, the agent had begun telling customers that moving their payment date would not affect interest, which was not true for one product.

Customers were pleased. The answer was wrong. The issue was caught by a quality review, not by CSAT, and the agent was dropped to read-only for that capability until it was fixed and recertified.

How is CSAT different from an answer quality score?

CSAT is the customer's view of the interaction. An answer quality score is the firm's assessment of whether the agent's answer was correct, grounded in approved sources, within the agent's role and handed over when required. It is measured by reviewers or by scoring an AI agent's answers against known cases.

You need both, and they answer different questions. CSAT tells you whether customers found the agent helpful. Quality scoring tells you whether the agent should be allowed to keep doing the job. Only the second belongs in certification, and even then an eval score is not a bound: it informs the decision, it does not set the limits.

Which terms sit next to CSAT?

• Customer effort score: how easy the interaction was, which often explains a low CSAT.
• Net promoter score: willingness to recommend, a relationship measure.
• Quality assurance: reviewing a sample of conversations against a standard.

Where Pryme fits

In Pryme, satisfaction is an operational signal, never certification evidence. A support agent that knows when to say no earns depth through Attest gates such as grounded, refusal and escalation, scored on the customer's own past cases. A high CSAT cannot lift an agent past a failed gate.