Empathy has long been the soft skill that everyone in customer experience preaches and almost nobody measures. A new analysis from Level AI suggests that era may be ending, and the findings should get the attention of anyone running a collection floor.
The company used its AI platform to analyze 35.1 million customer conversations across 85 companies this year, asking a simple question: does it matter when an agent explicitly acknowledges a customer’s frustration, stress, or concern? The answer was a clear yes. Conversations that included explicit emotional validation scored an average inferred customer satisfaction score of 3.94 on a five-point scale, compared with 3.52 for conversations without it, a lift of 11.84%.
Rather than relying on the small fraction of customers who respond to post-call surveys, Level AI’s inferred CSAT metric uses AI to score every interaction on sentiment, resolution, and customer effort, allowing the company to measure satisfaction across the full population of conversations.
The category-level results are where things get interesting for the accounts receivable management industry. Financial services and insurance, a category the report says includes debt conversations, disputes, claims, and payment issues, showed a 14.3% satisfaction lift when agents acknowledged customer emotion, the second-strongest result in the entire dataset behind only technology and SaaS at 14.8%. The report attributes the strength of those results to the uncertainty customers bring to those interactions, noting that acknowledgment carries more weight when the customer is dealing with a situation that already feels immediate.
That describes a large share of collection calls. Consumers contacting an agency about a debt are often stressed and dealing with a situation they would rather not be in. The data suggests that agents who name that emotion directly, rather than plowing straight into account details, are measurably improving outcomes.
Not every category saw the same benefit. In healthcare and wellness, where the report notes that coverage decisions and clinical constraints often sit outside an agent’s control, empathy produced only a 1.8% lift. Utilities and energy came in at 2.7%. Acknowledgment still helps, but it has less room to move the score when the underlying answer is out of the agent’s hands.
The findings align with broader consumer research. A 2025 Zurich survey found three in five customers would pay more to do business with a brand that demonstrates care, and a Five9 survey found 86% of consumers value empathy and human connection over a quick response.
Perhaps the most striking number in the report: only 18.2% of the 35.1 million conversations included explicit emotional validation at all. Four out of five interactions skipped it entirely. Level AI’s recommendation is to treat validation as an observable, scorable behavior, coached by category and measured across every conversation rather than a sampled handful. For agencies already deploying speech analytics and AI scoring tools, empathy may be the next behavior worth building into the rubric.
