Deloitte Finds Claims Customers Stay for Humans, Not Digital Tools
A new Deloitte analysis of nearly 4,000 auto and home policyholder responses across 12 property and casualty carriers lands on an uncomfortable finding for claims automation programs: the four strongest drivers of positive customer sentiment all turn on human interaction. Customer support, knowledge, communication, and attitude. Those four, the authors say, are the ones that influence future purchases and reinforce trust. The responses, collected by data firm HundredX, show policyholders still describing claims processes as complex, fragmented, and unclear even after large insurer investments in platforms and digital tools.
Every figure here, and the method behind it, reaches us through a single Carrier Management report; nothing under Deloitte’s own imprint turned up in a search. That report names four Deloitte authors and sizes the sample, but a reader who wants the study title, the fieldwork window, or the error range has nowhere to look it up. HundredX, whose data this is, sells the measurement. And a sentiment survey records what policyholders say drives loyalty, not what any of them did at renewal.
The eight variables the analysis weighed tell the story by what is missing. The claims process itself, service speed, claims outcome, and the app or website all matter, but none outrank the human variables. Sentiment weakens, the authors write, when a customer meets too many steps, slow processing, unclear processes or status updates, repeated information requests, or settlement delays. Deloitte does not attribute any of those to automation. Each is a point in the file where the customer is dealing with the process rather than with somebody who can answer for it, and the four variables that move sentiment most are the ones a process cannot supply on its own.
Deloitte’s prescription keeps AI in claims and changes where it sits: prioritize the highest-friction processes, streamline behind-the-scenes decision-making, keep humans at the center of high-emotion moments, route customers to the right channel, and broaden claims metrics to include customer retention. Only the last of the five items is about measurement; the other four decide where the AI goes, and all four put it upstream of the customer, leaving the customer-facing moment to a person. Whether that person changes any outcome is a different question, and a harder one to document.
That design is already law in places, though not in this line. Minnesota bars automation alone from prior authorization denials, and Utah limits the AI role in health preauthorization the same way. Both are health rules, and no property and casualty equivalent exists, so nothing in the Deloitte data creates an obligation for an auto or home claims operation. The exposure is evidentiary instead: channel-routing logic and human-review checkpoints acquire a second use once they are documented inside an AI governance program, where they become the record of where the handoff happened. An insurer whose claims AI inventory stops at the model and never records where the human takes over has nothing to show either the examiner or the customer.