Most customers say they'd rather you didn't use AI. Most also say they're open to it. Here's how both can be true, and what it means for your business.
Picture a head of customer service who likes the idea of an AI agent on her company's website but can't bring herself to switch it on. Her worry is simple: what if it says something wrong at 10pm and nobody is there to catch it?
It's one of the most common hesitations I hear, and the surveys suggest customers feel some of it too. Ask whether people trust AI customer service and you get two answers at once. A big share of customers say they'd rather companies didn't use it. A big share also say they're happy to use it, under the right conditions. If your company answers customer questions online, the useful part is in those conditions.
I build AI chat agents for a living, so I'm not neutral. That's why I'll stick to what the research found, including the parts that don't flatter my industry.
What the surveys actually say
Start with the skeptics. In a Gartner survey of 5,728 customers, run in December 2023 and published in July 2024, 64% said they would prefer companies didn't use AI in customer service. 53% said they would consider switching to a competitor if a company did.
Gartner also asked why. The top concern was that AI would make it harder to reach a person. Next came AI displacing jobs, and AI giving wrong answers.
Now the other side. A Five9 study of 1,006 consumers in the US, UK and Canada, published in March 2025, found 72% are open to AI-powered interactions, provided they can escalate to a human when needed. And 59% said they'd prefer an instant AI chatbot over waiting for a live agent.
A fair caveat: Five9 sells contact-centre software, AI included, and commissioned that survey through Zogby Analytics. Gartner's data is older and asked a broader question. Different questions, different years, different sponsors. Still, read together, they don't really contradict each other.
Why both numbers can be true
Look closely at the wording. The skeptics are mostly afraid of one thing: getting stuck with a bot and no way out. The open-minded group is open with one condition attached: they can reach a human.
That's the same fear, seen from both sides. People don't hate AI answering a question. They hate AI standing between them and a person when they need one.
Transparency matters too. In Salesforce's State of the AI Connected Customer report, published in July 2026 from a survey of 16,585 consumers and business buyers, 72% said it's important to know if they're talking to an AI agent. Customers don't want to be fooled. They want to know what they're dealing with, and they want a way out.
And business owners feel it from their side as well. A Simply Business survey of 1,047 US small business owners (Q4 2025 to Q1 2026) found 62% already use AI to run their business, but 86% rated the ability to speak to a human as important, with 66% calling it very important. Owners are using AI. They're just not handing it the keys.
Why trust in AI customer service matters for your business
A bad chatbot costs any company something, and when your revenue rests on a few dozen key accounts, it can cost a lot. One client who feels trapped by a bot mentions it to their account manager, or to a peer at another company, and you've lost more than one conversation.
So the manager in that picture is right to be careful. But "don't use AI at all" has a cost too. Picture a prospect on your site at 9:14pm asking whether your team can run a demo this week. If nobody answers until 9am, they've likely asked someone else. The research points to a middle path: let AI answer the easy questions fast, and make the route to a person short, visible and always open.
How we handle it: your team stays in control
When we built ChatX, this fear was the design brief. Here's what that looks like in practice:
- Take over, then Release to AI. Anyone on your team can take over a live conversation from the AtomicX mobile app or Slack, at any moment. When they're done, they release it back to the AI.
- Visible handoff. The agent hands off when it's stuck, when the visitor asks for a person, or when it spots a hot lead. You set the rules.
- Nobody repeats themselves. Your staff get the transcript, an AI summary, the reason for the handoff and suggested replies.
- Nothing runs without a click. Before the AI starts a workflow, the visitor has to click Start. Booking requests wait for your staff to Accept them. Store order access is read-only, so it can't issue refunds.
- You can check its work. Show AI Details lets you see the content and tool calls behind an answer, and Teach the agent turns your staff's better reply into the answer it gives next time.
None of this makes the AI perfect. It will sometimes get things wrong, and I won't pretend otherwise. The point is that a wrong answer is something your team can see, step in on and fix, not something buried in a chat log nobody reads.
The bigger pattern
Trust in AI customer service isn't really a question about AI. It's the same question customers have always asked about any business: will someone help me when it matters?
Businesses share the responsibility here. An AI agent with no escape route is a choice, and so is one that pretends to be human. Customers notice both. The surveys suggest they're willing to meet AI halfway, as long as a person is clearly within reach.
If you want AI answering fast but you're not ready to let it run alone, you don't have to. Want to see how Take over and Release to AI work on your own site? Book a 20-minute demo.