Pew asked 3,488 Americans about AI in healthcare. The human in the loop is the product
Published · By Gaurav Phogat
A Pew Research Center survey of 3,488 US adults, reported by Healthcare Dive on 4 November 2026, finds people comfortable with AI for tasks like booking appointments, but wary of it for diagnosis or claims, and much more trusting of providers than payers. My read for India: the visible human is not a cost to be automated away. It is the feature.
What happened
Healthcare Dive reported on 4 November 2026 on a Pew Research Center survey of 3,488 US adults. Respondents were comfortable with AI in back office work such as booking appointments and ambient scribing. About half thought AI could help translate medical jargon for patients, but only with a human reviewing the output; a quarter thought it could do so alone. 44 per cent thought AI could orchestrate scheduling with a human in the loop.
This article is part of the Industry Watch series, best read in order from the pillar.
Resistance rose for higher stakes uses. Four in 10 said AI should not make diagnoses even with human review, 60 per cent would be uncomfortable if their own provider used AI for a diagnosis, and 46 per cent said AI should not make insurance claims decisions even with a human involved. 58 per cent had little or no confidence that payers would use AI in patients’ interest, while providers fared better, with a quarter expressing a great deal of confidence and 34 per cent some. 51 per cent expected AI to hurt privacy and data security.
My take
This is a US survey, but the pattern matches what I see in Indian hospital digital journeys. People will happily let a bot find a slot, send a reminder or explain a report in plain language. What they do not want is to feel that a machine is making the call with no one accountable.
So the design brief for a hospital chatbot changes. The goal is not maximum containment. It is fast, useful automation for the simple jobs, with a clearly signposted human for anything that feels consequential. When I tested chatbots on 27 Indian hospital websites for my hospital chatbot study, the weak experiences were rarely about the AI itself. They were about dead ends.
What most coverage missed
The trust gap between providers and payers is a brand asset for hospitals. When patients trust their doctor more than their insurer, the hospital’s own front door, its app, website and WhatsApp, is where AI is most likely to be accepted. Handing that front door to an anonymous third party tool wastes the trust the hospital already has.
What I would do
- Automate booking, reminders, directions and report explanations first, where comfort is highest.
- Show the human: name the care team or desk behind the bot, and make handoff one tap away.
- Never let a patient facing bot appear to diagnose; route symptom questions to people.
- State plainly how patient data is used, since privacy worry is the majority view.
Source: Healthcare Dive. Figures as reported at the time of writing.
Sources and further reading
- Healthcare Dive (healthcaredive.com)
Questions people ask
As reported by Healthcare Dive, Pew surveyed 3,488 US adults. Respondents were comfortable with AI for admin tasks like booking, but many opposed AI diagnosis or claims decisions, even with human review.
Providers. 58 per cent had little or no confidence in payers using AI in their interest, while a quarter had a great deal of confidence in providers and 34 per cent some confidence.
Automation handles simple tasks such as booking and reminders, while a named person or desk reviews or takes over anything consequential, with an easy, visible handoff.
Not directly, since the health system and payers differ. But the underlying preference, automation for convenience with people accountable for decisions, is a sensible design principle for Indian hospital digital journeys too.
Read my takes first in Google Search
