The Support Exception
Why AI Support is Winning While AI Anxiety Rises
Executive summary
Most coverage of AI sentiment tracks how people feel about AI in the abstract, and that number is getting worse. This report asks a narrower question about AI in support specifically: how do people feel about AI when it is handling their support, and what do they actually want when they ask for a human. The answer diverges from the headlines.
2:1
Support is the exception to rising AI anxiety
Abstract AI feeling is souring while attitudes toward AI in support are improving. Among recent AI-support users, views of AI customer support over the past year ran 38.2% more positive against 19.3% more negative, close to 2 to 1 improving. That runs opposite to broad AI sentiment. Pew found 50% of Americans more concerned than excited about AI against 10% more excited, up from 37% concerned in 2021.
2.6x
Consumers name Amazon the best AI support, and quality tracks industry
Asked to name a brand with memorably good AI support, people named Amazon more than any other. Underneath that, the answers split cleanly along industry lines. Fintech and AI-native brands were named good 2.6x more often than bad. Telecom, utility, delivery, and courier brands were named bad 3.2x more often than good. The industries consumers trust most for AI support are the ones deploying it against complex, regulated work.
71%
The human preference is conditional
Most people are satisfied once AI solves the problem. Asked whether they’d still want a human if an AI resolved their issue correctly and quickly, 71.0% of all respondents said no, and only 9.5% would still want a human. The residual preference for a human is mostly doubt that AI will get it right.
69%
Good AI support builds the brand
An excellent AI support experience makes a brand more memorable. 69% of respondents said so, against only 3.8% who said it would make the brand less memorable. Getting AI support right is a brand asset in its own right.
47%
AI earns a second chance
When an AI fails on the first try, people split almost evenly between retrying the AI (47.3%) and asking for a human (48.1%). The moment after something goes wrong is a fresh chance to get it right, so recovery flows matter as much as first-try resolution.
Methodology
- Sample: 2,015 consumers
- Recent AI-support users: roughly 95% had interacted with an AI support agent in the last three months (only 5.1% reported none).
- Countries reported: United States, United Kingdom, and Canada
- Platform: Prolific, with respondent verification
- Data collected: July 2026
Finding 1: Support is the exception to rising AI anxiety
Support sentiment is improving about 2 to 1, against a falling general trend
Attitudes toward AI in support are improving even as broad AI sentiment falls. Asked how their opinion of AI customer support had changed over the past year, users landed here:
/ excited
/ concerned
That is close to two people improving for every one souring, and it runs against the broader direction of AI sentiment. Pew found 50% of Americans more concerned than excited about AI, against only 10% more excited, up from 37% concerned in 2021. Qualtrics XM Institute, surveying 24,000 consumers across 23 countries, found comfort using AI declined year over year.
Broad, abstract AI sentiment is getting more negative at the population level. The support-specific measure is improving among the people who actually use AI support. Support is where an abstract worry about AI meets a concrete outcome, and the concrete outcome is winning.
Finding 2: Consumers name Amazon the best AI support, and quality tracks industry
Amazon is named good more than any brand; underneath, quality splits by industry
We asked people to name a brand with memorably good AI support, and separately, memorably bad. Amazon was named good more than any other brand, because it is the AI support surface almost everyone has used.
AI support brand recall
2.4x
good
Underneath the single-brand noise, the answers split along industry lines:
AI support recall by industry
2.6x
good
delivery / courier 155 skews
3.2x
bad
The pattern holds regardless of brand size and reflects how far each industry has gotten with support automation that actually resolves issues. The industries consumers trust most for AI support are the ones deploying it against complex, regulated work.
Finding 3: The human preference is conditional
71% would be satisfied with AI once it solves the problem
We asked all 2,015 respondents a direct question: if an AI resolved your issue correctly and quickly, would you still want a human?
Once the outcome is good, roughly 7 in 10 stop caring who delivered it, and only about 1 in 10 hold out on principle. The stated preference for a human is largely a proxy for a fear that AI will not get it right. The wall of human preference is softer than a raw preference question suggests.
Finding 4: Good AI support builds the brand
69% say excellent AI support makes a brand more memorable
AI support makes your brand memorable
more memorable
A great AI support experience makes a brand more memorable. 69% of respondents said so (21.8% much more, 47.4% somewhat more), against only 3.8% who said it would make the brand less memorable. That reframes support automation from a cost line into a brand asset: getting it right is a way to be remembered.
Finding 5: AI earns a second chance
After a first-try failure, 47.3% retry the AI vs 48.1% ask for a human
When an AI fails on the first try, people split almost evenly between retrying the AI (47.3%) and asking for a human (48.1%). The moment after something goes wrong is a fresh chance to get it right, so recovery flows matter as much as first-try resolution.
What this looks like in practice
The teams already earning this trust are in the exact industries the data points to. The data describes consumer attitudes; here’s what happens when companies deploy AI in the industries this report says consumers trust most.
Fintech
Flex, which lets renters split monthly rent payments to avoid late fees, saw a 2x CSAT improvement over its previous support tool and a 50% drop in median time to resolution, holding up through the 4x chat-volume spike of rent week.
Matching urgency with accuracy at scale
Flex
Digital health
Eucalyptus, one of the largest GLP-1 telehealth clinics and now acquired by Hims & Hers, scaled patient support 4x without adding headcount with Lorikeet. Healthcare is complex, regulated work, exactly where consumers reward AI that resolves the issue.
Scaling patient support without adding headcount
Eucalyptus
“We tested AI solutions head-to-head and Lorikeet was a winner in every metric.”
Lindsay Boland · Customer Service AI Product Lead, Flex
Key implications
01
Support is a bright spot, so treat it like one.
Support sentiment improved close to 2 to 1 over the past year while broad AI sentiment fell. The window to build trust is open in support even as it closes elsewhere.
02
Deploy where the trust already is.
Consumers name Amazon the best AI support, and fintech and AI-native brands are named good 2.6x more than bad. The industries doing complex, regulated automation are the ones consumers reward.
03
Resolution is the whole game.
71% of all respondents are satisfied once AI resolves the issue correctly and quickly. Nail the outcome and the channel preference mostly disappears.
04
Good AI support builds brand.
69% say a great AI experience makes a brand more memorable. The upside runs past cost savings into brand recall.
05
Expect a second chance, then earn it.
After a failure, people are almost as likely to retry AI as to ask for a human. Recovery flows matter as much as first-try resolution.
Numbers that matter
2:1
38.2% vs 19.3% improving vs souring on AI support over the past year, against a falling general AI trend
Amazon
named the best AI support, more than any other brand
2.6x
fintech / AI-native named good more than bad
3.2x
telecom / utility / delivery named bad more than good
71.0%
would be satisfied with AI if it resolved the issue correctly and quickly
69%
say excellent AI support makes a brand more memorable
47.3%
retry the AI vs 48.1% who ask for a human after a first-try failure
What this means for teams deploying AI support
The through-line is that consumers are not anti-AI where it counts, and support is pulling away from the broader anxiety about AI. The human preference is conditional and emotional, the industries deploying AI against complex, regulated work are the ones consumers already trust, and getting the experience right builds brand. The teams that win here resolve issues correctly and quickly and use AI support to build brand.