Thursday November 12, 2026
Larassatti D.

In case item bankrupt on your website, would you let an AI delegate fix it?
If your archetypal intuition is to hesitate, you’re not alone. In surveys, lone a small portion of group say they’d let AI create changes on their behalf, equal alongside their approval.
But surveys ask concerning AI in broad and concerning situations that haven’t happened yet. We wanted to see what group do in genuine real-life situations. So we looked at how our customers worked alongside Hostinger’s AI assistance agent.
The delegate is built into the Hostinger dashboard, so customers didn’t have to search it out, opt in, or set item up. It’s merely how assistance works. That method the group in our data aren’t first adopters who sought out AI. They’re mundane customers who needed assistance alongside something.
Whenever customers asked for assistance alongside their hosting, domains, or email, it could inspect their accounts and acknowledge issues. If fixing a issue required the delegate to execute a change, it would continually ask for the customer’s endorsement first.
Our data shows that customers stated yes concerning 9 in 10 times (93%) whenever the delegate asked to create a change. This demonstrates a distinct image from the one surveys coat of the AI rely gap, the extend between using AI and trusting it to act.
What surveys say concerning the AI rely gap
Most of what we cognize concerning rely in AI comes from surveys, and they display a apparent tension. A earth study by the University of Melbourne and KPMG established that 66% of group use AI regularly, but lone 46% are consenting to rely it.
Trust drops additional whenever AI moves from assisting to acting. Capgemini established that the portion of executives who rely completely autonomous AI agents dropped from 43% to 27% in a sole year. In Ping Identity’s 2026 survey of 11,000 consumers, fair 5% would go so far as to let AI act on their behalf alongside their approval, and lone 3% would let it act without asking.
These surveys measure item distinct from what our data do. They ask how group awareness concerning AI in general. Our AI assistance delegate asked concerning one particular fix for a issue the client had fair described.
That difference is the point: our data shows what rely looks akin whenever the petition is genuine and specific.
Our AI assistance agent’s project delegation funnel
Most assistance questions didn’t need any changes. But whenever one did, our AI assistance delegate went through three steps:
- Looked – In additional than 1 in 4 conversations (29%), our AI assistance delegate checked the customer’s document to see what was going on.
- Asked – In concerning 1 in 5 of those conversations (20%), our AI assistance delegate established item value changing and asked for approval to obtain action.
- Approved – Customers stated yes to 93% of those confirmation requests.

Month by month, customers approved between 89% and 95% of the changes the AI assistance delegate requested from October 2025 to August 2026. If a client didn’t verify a alter inside 30 minutes, we counted it as not approved, despite of the reason.
There’s a parallel in a extremely distinct product. Anthropic reports that developers using its coding tool, Claude Code, accept 93% of manual approval prompts. The matching figure is a coincidence, and the settings are different, but the form is similar: whenever an delegate asks before acting, group mostly say yes.
What are the tasks customers hand complete to our AI assistance agent?
Customers mostly hand complete the behind-the-scenes setup and upkeep that keeps a website online. DNS and domains came out on top, at 34% of all alter requests. These settings decide anywhere your website and email go, and one incorrect admission can obtain the two offline.
Here’s how the top five areas compare:
- DNS and domains – 34%
- WordPress – 14%
- Servers – 14%
- Security and SSL – 10%
- Speed and caching – 9%

Approval stayed elevated throughout the board, from 94% for DNS and domains downward to 90% for speed and caching. Even the lowest area motionless got a yes 9 times out of 10.
Updating DNS records, clearing WordPress cache, restoring backups, installing SSL certificates, and deploying apps on a server are the actions most delegated to our AI assistant.
Approval barely changed by category of action
Among the tasks handed complete to our AI assistance agent, deletions were approved as frequently as creating or updating something.
From the riskiest-sounding act to the safest, the entire range is fair three percent points. Whether the aide wanted to delete, create, reset, or rotate item off, customers stated yes at nearly the identical rate.

That’s not how rely looks in broad AI delegate use surveys, anywhere it drops as the stakes go up. In the agentic AI statistics we’ve gathered, for example, lone 20% of executives rely agents alongside financial transactions, compared alongside 38% for data analysis.
Our data doesn’t shield money-related decisions, though. Billing, refunds, and payments weren’t part of this study, so these results are lone concerning specialized fixes.
When an delegate handles payments or endeavor decisions, the stakes and the risks appearance different. Our guide to AI delegate risks covers what can go incorrect and how to remain in control.
How accelerated did alter requests grow?
Faster than conversations did. Between the archetypal five months of the study (October 2025 – February 2026) and the final five (April – August 2026), conversations grew concerning 1.7x, and alter requests outpaced that in six of the eight areas.
Routine website upkeep led the way: speed and caching changes grew concerning 2.5x, and WordPress changes concerning 2.4x.

To keep the difference fair, we counted lone tools that already handled alter requests in October 2025 and excluded March 2026 to forestall the two periods from overlapping.
A quick term of caution: the data shows growth, not the decisions rearward it. Customers may have been handing complete more, or the delegate may have been suggesting changes additional frequently as it improved.
What we can say for certain is that additional requests didn’t bring additional hesitation. Approval held between concerning 89% and 95% the entire way through.
The new AI delegate skills inform a akin story
The difference complete omits the new skills the delegate gained in 2026, and customers began using them as shortly as they were launched. In AI agents, a accomplishment is a particular project the delegate can execute on its own. Each new accomplishment expands what the delegate can do for our customers. Installing WordPress, accessible from April, is one example.
In April – August 2026, new skills made up additional than 40% of website-building and email requests. That additionally explains why website construction and email were the two areas that didn’t outpace conversations on the first tools: much of that activity had moved to the new ones.
Once group delegate, they keep going
A archetypal yes isn’t continually the last. Between June and August 2026, in 1 in 4 conversations (25%) anywhere customers approved a change, they went on to hand complete additional tasks in the identical conversation.
That fits a broader trend: rely in AI tends to develop the additional group use it. In Ping Identity’s 2026 survey, 43% of consumers stated they rely AI additional today than a twelvemonth ago. Our data shows a smaller type of the identical pattern, playing out inner a sole conversation.
How asking approval closes the AI rely gap
Hostinger’s AI assistance delegate continually asked for confirmation before making any alter to a customer’s account, and it lone needed to ask in concerning 1 in 18 conversations. When it did, customers stated yes 9 out of 10 times.
Unlike the surveys, which asked concerning AI in general, our customers were deciding on particular tasks, and without their yes, the alter didn’t happen.
That’s what individual in the iteration looks akin in practice: the aide does the legwork, and the client makes the call.
Surveys item the identical way whenever the inquiry gets specific. In Accenture’s study of 25,590 consumers throughout 16 countries, 74% would let an AI delegate grip regular tasks specified as resolving complaints, as lengthy as it strictly followed their instructions.
Gartner additionally predicts that agentic AI volition determine 80% of average client assistance issues without individual assistance by 2029. If that’s anywhere things are heading, our data suggests achievement volition depend on involving customers in decisions at the correct moment.
Our AI assistance delegate has now merged into a sole tool alongside Hostinger Agent, an agentic merchandise that handles endeavor tasks akin promotion planning, satisfied writing, and SEO. Help alongside hosting, domains, and email is included alongside all hosting plan, during the paid Agent Plus scheme lets users hand complete additional of the activity of operating their business.
The two tools covered a lot of the identical ground, so we brought them together to create operating a website equal easier. Ričardas Kudirka, Hostinger Agent’s Product Manager, explains what connects them: “Trust doesn’t arrive from the AI acting on its own. It comes from the AI being apparent and the client making the decision. That’s the regulation we’re construction Hostinger Agent on.”
Methodology
This study is according to anonymized, aggregated use records from Hostinger’s AI assistance agent, covering October 2025 to August 2026, previous to its integration into Hostinger Agent on September 2, 2026.
We excluded September 2025 since we were motionless construction out the agent’s tools. We additionally excluded conversations concerning billing, refunds, and payments.
The records display which tools the aide used, whether a tool changes something, approval requests, confirmations, and timestamps. They merge no communication content, names, domains, or document IDs. No conversation satisfied or prompts were analyzed.
Key definitions:
- Looked: the aide checked the customer’s document at smallest formerly in a conversation.
- Asked: the aide projected a alter that needed the customer’s permission. Every specified alter required confirmation.
- Approved: the client confirmed the alter inside 30 minutes. A petition that wasn’t confirmed in that opening didn’t go onward and is counted as not approved, so endorsement rates are conservative.
Growth figures difference April–August 2026 alongside October 2025 – February 2026 and contain lone tools that already handled alter requests in October 2025.
These data do not connect to customers’ use of Hostinger Agent, as no Hostinger Agent conversations were analyzed for this study.
What the data can’t inform us
These figures arrive from customers who had already asked for help, not from the broad public. They display what customers approved, but not why. Some of the growth in alter requests may indicate the assistant’s improving tools fairly than changes in client habits. People may rely an delegate alternatively alongside a assistance fix than alongside a acquisition or a endeavor decision.