How do customers feel about AI chatbots? Mostly wary. In a Gartner survey of 5,728 customers, 64% said they would prefer companies didn't use AI for customer service. The reasons behind that number are more useful than the number itself, because most of them are things a business can fix.
I'm the founder of JIVIQ, which makes an AI chat assistant for websites, so I have a stake in this question. I've kept the research claims to sources you can check, listed at the end.
What the research says
Gartner surveyed 5,728 customers in December 2023 and published the results in July 2024. Two findings stand out:
- 64% would prefer that companies didn't use AI for customer service.
- 53% would consider switching to a competitor if they found out a company was going to use AI for customer service.
The top concern was that it would get harder to reach a person. After that came worries about AI displacing jobs and AI giving wrong answers.
Gartner's own advice in the release was that AI chatbots should tell customers they will connect them to a person if the AI can't solve the problem. In other words, the fear is less about AI itself and more about being trapped with it.
Why older chatbots earned a bad reputation
Rule-based chatbots could only handle questions they had been scripted for. If a customer phrased something differently, or asked two things at once, the bot fell back to a menu.
Customer:
"I can't access my account because
I forgot my password and changed my email."
Bot:
"Please select one of the following
options..."
After enough of these, people start a chat expecting it to fail. That expectation carries over to newer tools, even when they work better.
Concern 1: "I won't be able to reach a person"
This was the top worry in the Gartner survey, and it is the easiest to address. Tell people up front that a person is available, and make the handoff work.
In JIVIQ, for example, the assistant introduces itself as an AI and says "Want a person instead? Just ask." Visitors can ask for a human at any point, and when a teammate replies, the AI steps back for that conversation. Whatever tool you use, check that the path to a person is visible and that someone is actually notified when it is used.
Concern 2: "The AI will give me a wrong answer"
This one has real consequences. In Moffatt v. Air Canada (2024), a Canadian tribunal held the airline responsible after its website chatbot gave a customer incorrect information about bereavement fares. The tribunal rejected the argument that the chatbot was responsible for its own statements.
The practical lesson: an AI assistant on your website speaks for your business. It should answer from your own approved content, and it should say when it doesn't know rather than guess.
That's the approach we took with JIVIQ. The assistant only answers from the knowledge base you give it (your website pages and uploaded documents), checks its draft against those sources before sending, and, when the answer isn't there, says "I couldn't find that in the knowledge base, so I won't guess" and offers to pass the chat to your team.
Concern 3: "I didn't know I was talking to a bot"
The Gartner switching figure is about finding out a company uses AI. Hiding it makes that worse. Label AI replies as AI. In JIVIQ, AI messages carry an AI badge, and the assistant names itself as your business's AI assistant at the start.
What customers still want: fast, short, correct
None of this means customers reject quick automated answers. A customer asking for your opening hours or refund window usually just wants the answer. A good reply is:
- Short and easy to read
- Specific to the question asked
- Taken from your own policies, not general web knowledge
- Followed by an easy way to reach a person if it isn't enough
A three-paragraph answer to a one-line question feels like a wall, even when it's correct.
Where people should take over
AI handles repeated questions about products, pricing and policies well when those answers are written down. Billing disputes, unusual requests, complaints and anything sensitive are better handled by a person. Decide in advance which is which, and set up your tool so the handoff happens without the customer having to fight for it.
In JIVIQ you can choose how much the AI does per workspace: Autopilot (AI answers, hands off when it can't), Copilot (AI drafts, a person approves each reply) or Off (people only). Copilot is a reasonable start if you want to see the AI's answers before customers do.
A short checklist before you turn on an AI chatbot
- Does the bot say it is an AI?
- Can a visitor reach a person in one message?
- Does someone on your team get notified when that happens?
- Does the bot answer only from your own content?
- Does it say "I don't know" instead of guessing?
- Have you read a week of its conversations yourself?
Final thoughts
The research says many customers are wary of AI in customer service, mainly because they fear losing access to a person and getting wrong answers. Both are design choices. Be open that it's AI, keep it to your own content, and make the handoff easy.
If you want to try that setup on your own site, JIVIQ has a 14-day free trial with no card required.
How we checked this
- Gartner, "Gartner Survey Finds 64% of Customers Would Prefer That Companies Didn't Use AI For Customer Service," July 9, 2024: gartner.com press release
- Moffatt v. Air Canada, 2024 BCCRT 149, Civil Resolution Tribunal of British Columbia: published decision
- JIVIQ behaviour described here comes from our help center.