Do 64% of Customers Really Not Want AI in Customer Service? What Gartner Asked (2026)

The 64% is real, but it is not what most retellings say. In a Gartner survey of 5,728 customers fielded in December 2023 and published on 9 July 2024, 64% said they would prefer that companies didn't use AI for customer service, and 53% said they would consider switching to a competitor if they found out a company was going to use it. It is a stated preference about a hypothetical, not a refusal to talk to AI. In a separate question in the same survey (n = 5,140), the top concern, named by 60%, was that AI will make it harder to reach a person.

This page traces the number back to Gartner's own release, sets it beside the other customer surveys and studies we could verify, with their samples, dates and (where published) exact questions, shows how "would prefer" became "hate" within four days, and lists what Gartner never said. It is the same kind of source audit we ran on every lead response time study. For how many companies already use AI in service, see our AI customer service adoption statistics.

Do 64% of customers really not want AI in customer service?

Partly. This is Gartner's own sentence, word for word: "Sixty-four percent of customers would prefer that companies didn’t use artificial intelligence (AI) in their customer service, according to a survey by Gartner, Inc. Furthermore, 53% of customers would consider switching to a competitor if they found out a company was going to use AI for customer service." (Gartner, 9 July 2024)

Three things about that sentence matter more than the number itself:

1. It is a preference about a hypothetical. "Would prefer", "would consider switching", "if they found out a company was going to use AI". It does not measure what people did when an AI actually answered them.
2. It is about AI in customer service in general. The release names no channel. Where it gives an example, the example is a chatbot, not a phone call.
3. It came with a separate question on concerns. The same release carries a chart of customers' concerns, from a separate question with its own base (n = 5,140), and the top one is "It will become more difficult to reach a person" (60%).

The defensible reading is this: in late 2023, most customers asked in the abstract said they would rather companies did not use AI in service, and the top concern the same survey recorded was losing the route to a person. "64% of customers hate AI" or "64% won't talk to AI" is not what was measured.

What exactly did Gartner ask, and who did it ask?

Two cautions follow from what is missing. First, do not write "64% of 5,728". The headline sample is 5,728, but Gartner does not say how many people answered the 64% item, and the only chart in the release uses a smaller base of 5,140. "In a Gartner survey of 5,728 customers" is the accurate phrasing. Second, do not add a country, a channel or a question wording. Gartner has not made those details public, so an article that states them is not quoting Gartner.

Why do customers worry about AI customer service?

The same release answers this, in a chart most retellings skip. Asked "What are your major concerns related to AI customer service?", customers named:

Gartner's text puts it the same way: "The top concern that consumers have about AI in customer service is that it will get more difficult reaching a person (see Figure 1), followed by AI displacing jobs, and AI providing the wrong answers." (Gartner, 2024)

One trap: the release contains two different 60% figures. 60% of customers fear it will get harder to reach a person. Separately, Gartner's analyst says 60% of customer service and support leaders are under pressure to adopt AI. Different groups, different questions. Do not merge them.

Is the fear rational?

On the evidence, yes. From the same December 2023 survey of 5,728 customers, Gartner reported in August 2024 that only 14% of customer service issues are fully resolved in self-service, and only 36% of issues that customers describe as "very simple", although 73% of customers use self-service at some point. 45% of customers who started in self-service said the company didn't understand what they were trying to do. We cite the same 14% figure in our AI receptionist statistics and in our AI receptionist trends for 2027.

A US regulator had already documented the same pattern. In June 2023 the Consumer Financial Protection Bureau wrote that "Automated responses by a chatbot may fail to resolve a customer’s issue and instead lead them in continuous loops of repetitive, unhelpful jargon or legalese without an offramp to a human customer service representative." It called these "doom loops", and estimated that about 37% of the US population interacted with a bank's chatbot in 2022 (CFPB, Chatbots in consumer finance, 2023).

So the top concern in the survey is specific. Customers expected AI to stand between them and a person, and in 2023 that is often what self-service did.

“Many customers fear that GenAI will simply become another obstacle between them and an agent. The onus is on service and support leaders to show customers that AI can streamline the service experience.”

How did "would prefer" become "64% hate AI"?

Within days of the release, tech headlines turned a stated preference into hostility. The table traces each version of the number to the earliest place we found it. These are the earliest we found, not proven firsts.

Two patterns repeat. The verb hardens: "would prefer" becomes "hate", "repulsed" or "at all". And the condition drops: "if they found out a company was going to use AI" becomes "dealing with AI", or a channel the release never named. Each step makes the number sound like a measured refusal, which it never was. The Insurance-Canada.ca reprint is the exception, and a useful one: it is a word-for-word copy of the release, so citing it is citing Gartner.

Other 64% figures that get mixed up with Gartner's

At least four other 64% figures circulate on the same topic, and some point the opposite way.

How numbers drift inside one publisher

Recycling happens inside a single source too. SurveyMonkey's February 2026 roundup says "four in five (79%) Americans strongly prefer interacting with a human over an AI agent", from a December 2025 study of 2,017 US adults, but the same sentence gives a split of 70% for women and 60% for men, which cannot produce 79%. The same page labels as "2025" an "8% prefer AI" finding whose reasons (41%, 37%, 30%) are identical to its April 2023 poll, and its summary list (63%, 84%, 89%) matches Kinsta's 2025 figures (62.7%, 84.0%, 88.8%). We rate that 79% Low and keep it out of the study table below.

What did Gartner not say about the 64%?

Each of these was checked against the full text of the release.

1. Gartner never said 64% "hate", "won't talk to" or "refuse" AI. The release says "would prefer that companies didn’t use" AI. "Hate" appears in a TechSpot headline the next day; "repulsed" in Futurism four days later.
2. Gartner never said the 53% would switch "if they are dealing with AI assistants on the phone". The phone wording is TechRadar's. The release covers AI for customer service generally, and its example is chatbots. No channel split is published.
3. The question wording for the 64% and the 53% is not published, and neither are the countries, the survey mode, or whether respondents had used AI customer service. Only the concerns question is.
4. The base for the 64% is not stated. 5,728 is the survey; the concerns chart uses 5,140.
5. The 64% is not a 2025 or 2026 number. Gartner's July 2026 customer survey release does not repeat it. We searched its text for "64", "prefer" and "switch" and found none.
6. No study we found shows that people who try a well-designed AI agent with a clear handoff stop preferring humans. The post-use data that exists is poor for chatbots (details below).
7. The "64% say a human understands my needs better" figure is not Gartner's. It appears in TechRadar's write-up of an Okta survey, and not on Okta's own summary page.

What do other surveys say about AI vs human customer service?

Every head-to-head human-versus-AI preference survey we could verify and list in the table below favours people. But samples, sponsors and questions differ, and almost none publish the exact question. Here is every study we used, with its dates, sample, exact question where published, and our confidence.

How to read the table:

• Most of the strongest numbers are vendor-commissioned. Five9, Kinsta, Okta, Twilio and Liveops all paid for or ran their own research. That does not make them wrong, but it is why they are rated Medium or lower. Only Gartner's surveys and the peer-reviewed field experiment are rated High.
• The questions are not the same. Five9's 75% is a channel preference: "prefer talking to a real human in-person or over the phone for customer support". Okta's 70% is about AI agents in general, not customer service only. No survey in the table publishes the question behind its headline preference figure.
• Age matters. In Five9's data, 66% of Gen Z prefer a human for customer support, against 76% of Gen X and 86% of Baby Boomers and the Silent Generation.
• The objection is to AI with no way out. Kinsta's own reading of its survey is: "Respondents don’t dislike AI per se, but they dislike AI as a sole provider of customer service." The same survey found 88.8% think companies should always offer the option to speak with a human.

Do customers still want to talk to a human when AI answers?

Yes, and this is the most consistent finding in the whole set. Whatever customers think of AI itself, most of them want a person to stay reachable:

• Gartner, 2026: in a survey of 3,566 B2B and B2C customers in February and March 2026, 87% of customers expect an option to reach a human when using AI (Gartner's July 2026 chart). The same release found customers are about three times more likely to use third-party GenAI tools than company-provided chatbots, and that use of company chatbots "has remained statistically unchanged since 2022", while use of third-party GenAI tools in service "has nearly doubled in the past year". CMSWire reported the same findings.
• Kinsta, 2025: 88.8% think companies should always offer the option to speak with a human.
• Twilio, 2025: 78% of consumers say it is critical to be able to switch from an AI agent to a human when needed.
• Okta, 2025: 38% of respondents believe human oversight to review or approve AI agent decisions would increase trust.
• Liveops, 2025 (Low to Medium confidence): 55% of 1,000 US holiday shoppers had to escalate an AI-handled issue to a human.

Read together, the surveys say the same thing: most customers want a person to stay reachable when AI answers. That is exactly what the 60% in Gartner's 2024 chart were afraid of losing. To hear what a voice agent sounds like before you decide, our page can AI really talk like a human? lets you try one live.

Does trying AI make customers prefer it?

Not on the evidence we found, and it is worth saying plainly because the claim is often implied. We looked for studies that measure preference after people use a well-designed AI agent with a clear handoff, and found none. The post-use data that does exist is poor for chatbots:

• Gartner, 2023: in a survey of 497 B2B and B2C customers (December 2022 to February 2023), only 8% had used a chatbot in their most recent customer service interaction, and of those just 25% said they would use that chatbot again. Only 17% of billing disputes were resolved by customers who used a chatbot at some stage, against up to 58% for returns or cancellations.
• SurveyMonkey, April 2023: Net Promoter Score was +6 after a customer service interaction with a human and -66 after one with an AI.
• Liveops, 2025: 29% said AI improved their experience and 14% said it made it worse.

The nearest evidence of a shift is behavioural, not stated. In a randomized field experiment with more than 6,200 customers receiving outbound sales calls (Luo, Tong, Fang and Qu, Marketing Science, 2019), undisclosed chatbots were as effective as proficient workers and four times more effective than inexperienced workers at producing purchases. But telling customers they were talking to a chatbot before the conversation reduced purchase rates by more than 79.7% (a relative drop, not percentage points). The authors attribute this to "a subjective human perception against machines, despite the objective competence of AI chatbots", and found the effect smaller with late disclosure and with customers' prior AI experience.

This is not an argument for hiding AI. Many jurisdictions require callers to be told they are speaking to an AI (see our AI caller disclosure laws by country), and telling them is the honest default. What the experiment shows is narrower: part of the reaction to the word "AI" is a perception, and in that experiment it was smaller among customers with prior AI experience. The honest conclusion from all of it: acceptance follows task fit and a working handoff, not exposure alone.

Is there a law that gives callers the right to talk to a human?

In Spain, yes. Ley 10/2025 on customer service, published in the Boletín Oficial del Estado on 27 December 2025 and in force since 28 December 2025, says (our translation of the Spanish text):

• No bot-only service. Article 8.1 bans automated answering systems as the only channel: "Se prohíbe el empleo de contestadores automáticos u otros medios análogos como medio exclusivo de atención a la clientela."
• A person on request, from the start. Under Article 8.2, answering machines, conversational bots and similar systems must let the customer ask for personal attention from the main menu, at any moment of the interaction, from its start.
• Fast. Personal attention means a person answering in real time, and 95% of requests for it must be answered, on average, in under three minutes (Article 8.2).
• Twelve months to adapt. The transitional provision gives companies in scope twelve months from entry into force.

Who is covered is set in Article 2 of the BOE text. First, every company that provides basic services of general interest in Spain, whatever its size: water, gas and electricity supply; passenger air, rail, sea and bus or coach transport; postal services; electronic communications; and financial services (Article 2.1). Second, any other company or group selling goods or services in Spain mainly to consumers that, in the previous financial year, had at least 250 workers, or annual turnover above 50 million euros, or a balance sheet above 43 million euros (Article 2.2). Any one of the three is enough. A business outside the basic-service sectors and under all three thresholds is not covered. For financial services and electronic communications, their own sector rules come first and this law applies as a supplement, and some of its articles do not apply to the financial sector at all (Articles 2.5 and 2.6).

Beyond Spain, Gartner predicted in February 2024 that "By 2028, the EU will mandate “the right to talk to a human” in customer service interactions", in response to "growing consumer backlash against AI-generated service" (Gartner, 19 February 2024). That is a forecast, not a law, and this page makes no claim about EU-level rules. Whether an AI must say it is an AI is a separate question, covered in our country guide to AI caller disclosure.

What makes customers accept AI in customer service?

Four conditions come up again and again in the data:

1. It resolves faster. 72% of consumers in Twilio's survey say they would gladly use an AI agent if it could guarantee a faster resolution than a human. Gartner's 2023 survey found the chatbot's ability to move the customer's issue forward was the top driver of adoption. Among the 8% in SurveyMonkey's 2023 poll who prefer AI, the reasons were better availability (41%), speed (37%) and more accurate information (30%).
2. The task fits. In the same SurveyMonkey poll, more than half would be comfortable using AI to help with ordering food or beverages (65%), returning an item (59%) or completing work or school tasks (52%). At the other end, Twilio found consumers trust human agents for high-stakes issues: help with medical issues (64%, a different 64% from Gartner's), making an insurance claim (59%), handling returns or refunds (48%) and billing questions (45%). Returns sit in the middle, and the two surveys asked different questions.
3. It can act, not only answer. In Gartner's 2026 survey, 58% of customers who use GenAI said they have used it to complete a task on their behalf, rising to 74% in B2B.
4. The handoff keeps the context. In Twilio's survey, only 15% of customers feel the human has full context from the AI conversation when it moves to a person. Gartner's 2024 advice is that a chatbot must tell customers it will connect them to an agent if the AI cannot solve the problem, and "must then seamlessly transform into an agent chat that picks up where the chatbot left off."

“Many company-provided chatbots are still designed primarily to answer questions, but customers increasingly expect AI to help them take action, such as booking an appointment, submitting documents or updating their account.”

What is the reliable figure to quote in 2026?

If you need a current number, use Gartner's newest customer data instead: in its survey of 3,566 B2B and B2C customers in February and March 2026, 87% of customers expect an option to reach a human when using AI (Gartner's July 2026 chart). It speaks to the top concern in the 2024 survey, reaching a person, from a survey two years newer.

Avoid: "64% hate AI", "64% won't talk to AI", "53% would rather switch", anything "on the phone", "64% of 5,728", and the 64% presented as a 2025 or 2026 figure. And never put Gartner's 64% next to the other 64% figures above as if they agreed.

What does this mean for a business adding AI to its phones or chat?

It does not mean "don't use AI". Read honestly, the data says don't put AI between customers and a person. The top concern in Gartner's survey was that it will become more difficult to reach a person (60%). Kinsta's reading of its own survey is that respondents dislike AI "as a sole provider of customer service". And in Gartner's 2026 survey, 87% expect an option to reach a human when using AI. The design that fits that evidence is a capacity model, not a replacement one: the team keeps serving customers, and AI adds the capacity it lacks.

• AI takes the volume the team can't get to: the overflow when every line is busy, the busy hours, evenings and weekends. Among the 8% in SurveyMonkey's 2023 poll who prefer AI, better availability was the most-cited reason (41%).
• A caller can always ask for a person: transferred with a summary when someone is free, a message or callback booked when not. Gartner's analyst says customers must know the AI will connect them to a person when necessary; for companies in scope in Spain the law goes further: a person on request at any moment, from the start of the interaction.
• The work lands done. Point AI at tasks it can finish, such as bookings, order and status questions and simple changes, and have the result recorded in the CRM so nobody has to redo it. Gartner's 2023 survey found a chatbot's ability to move the issue forward was the top driver of adoption. Our guide to what to automate in customer service goes through the split.
• People take what needs them, with a summary, so the customer never starts again. Keep high-stakes and emotional matters with people. In Twilio's survey, only 15% of customers feel the human has full context after a handoff, so this is the step most worth getting right.
• Tell callers they are speaking to an AI, where the law requires it and as a default.

For the practical steps, see how to automate customer service without losing the human touch. AInora's voice agents are built this way: the agent picks up the calls your team can't get to, the work it finishes lands in your CRM, and when a call needs a person it transfers it with a summary of who is calling, what they need and what has already been discussed. When nobody is free, it takes a message or books a callback, so a caller who wants a person is never stuck with the AI. The same team serves more customers. To see how this would work on your own calls, tell us what your calls look like.