
Most talk about AI support agents is about how much they can do on their own. On 12 November 2026, web host Hostinger published a number that points the other way. Its AI support agent always asks the customer before changing anything, and customers said yes to those requests 93% of the time.
That figure is worth a closer look, and so is how it was measured.
The design rule, in Hostinger's words:
"If fixing a problem required the agent to perform a change, it would always ask for the customer's approval first."
And the result:
"Our data shows that customers said yes about 9 in 10 times (93%) when the agent asked to make a change."
Two details give the number its scale. The agent only needed to ask in about 1 in 18 conversations, and, in Hostinger's words, "Month by month, customers approved between 89% and 95% of the changes the AI support agent requested from October 2025 to August 2026."
Ričardas Kudirka, Hostinger Agent's product manager, puts the takeaway this way: "Trust doesn't come from the AI acting on its own. It comes from the AI being clear and the customer making the decision." Or, in the article's own phrasing: "That's what human in the loop looks like in practice: the assistant does the legwork, and the customer makes the call."
The sentence most readers will skip is the one that makes the number credible:
"If a customer didn't confirm a change within 30 minutes, we counted it as not approved, regardless of the reason."
Think about who falls into that bucket. A customer who started a chat, got called into a meeting and came back an hour later. Someone whose browser tab closed. Someone who meant to click "yes" and got distracted. All of them are counted as refusals.
So the real share of customers who would have approved is at least 93%, and probably somewhat higher. Hostinger chose the measurement that makes its own number look worse, which is the opposite of what most vendor statistics do, and it is a good reason to take the figure seriously.
One number in the same article is not Hostinger's: the forecast that AI "will resolve 80% of common customer service issues without human help by 2029" is Gartner's prediction, which Hostinger cites. It is a forecast, not a measurement.
We build DeskLeap, a helpdesk with ticketing, live chat and an optional AI Agent, so we checked our own code against these questions on 13 November 2026.
DeskLeap's AI Agent does not show the customer an approval prompt before it runs an action. What controls it instead is set by the business:
That is a different safeguard from asking the customer, and Hostinger's data is a fair argument that a customer-facing confirmation step is worth having for actions that change something. If the actions you would give an AI include refunds, cancellations or account changes, that is a good question to ask of any vendor, us included.
Related: Zendesk's AI agent launchpad and what you approve, on approval at setup time rather than in the conversation.