Customer service experience: what customers notice

Customer service experience is what getting help from you is like, seen from the customer’s side. Customers notice five things. They notice how long they wait, whether the answer is right, whether they can reach a person, whether they have to repeat themselves and whether you answer in their language. None of the five needs a support team. The last section shows how to check yours.

Deacon is an AI customer support agent for founders who answer support themselves. It answers your customers in seconds, from your own docs, in a chat on your website, and hands you the questions it can’t answer. It also shows you what customers are stuck on and what they want next. There’s a free plan, and it needs no card.

Every figure below links to its source, and each was read there on 27 September 2026.

How long customers wait for an answer

Customers expect an answer whenever they ask, nights and weekends included. In Zendesk’s CX Trends 2026 survey of 6,182 consumers in 22 countries, run in June 2025, 74% said they now expect customer service to be available around the clock because of AI.

Software support teams already reply fast. Freshworks’ 2026 benchmark, drawn from 33,889 accounts on its customer service products between June 2025 and May 2026, puts the median first response time in software and internet at 30 minutes 11 seconds. The mark for the fastest 20% was 21.6 seconds.

For a founder, the wait that hurts is the one while you’re asleep or building. A customer who writes at 11pm and hears back at 9am waited ten hours, however your helpdesk counts it. Time your waits in calendar hours, and first response time shows how to work them out.

Deacon answers in seconds at any hour, so a customer who asks at 2am about something your docs cover gets the answer at 2am. You hear quickly when a visitor is waiting for you. The owners on your team get an email three minutes after a visitor leaves an address or asks for a person. If someone on your team already has the dashboard open, a toast or the unread badge tells them instead. Analytics › Widget keeps a count of everyone still waiting for a reply from you.

Whether the answer is right

A fast wrong answer is worse than a slow right one, because the customer acts on it. Gartner surveyed 3,566 B2B and B2C customers in February and March 2026, and only 27% said they’d be willing to try a chatbot again after a negative experience. Gartner says customers whose chatbot misunderstood their issue, gave generic information or made it harder to reach a person are more likely to pick another channel next time. Its advice to service leaders is to “prioritize reliability over reach”.

A right answer comes from your own docs and matches the product as it is today. It answers the question the customer asked, not a nearby one. And when your docs don’t cover the question, the reply says so rather than inventing something. A bot that makes up a refund policy costs you more than the question would have. Why support bots make things up explains how that happens.

Deacon answers from your own content, meaning your website, your files, text you paste in and the answers you write. When those don’t cover a question, it says it doesn’t know instead of guessing, and asks for the visitor’s email. An answer taken from a page on your site usually links that page, so the customer can read on.

The gaps close as you work through them. Every question your docs didn’t cover waits in Conversations with an Answer this button. What you write becomes part of what Deacon knows, so the next customer who asks can get the right answer. Save & check shows you the reply Deacon now gives, and the more you answer, the less Deacon needs you.

Whether they can reach a person

Knowing they can reach a person is what makes customers willing to try everything else. In Gartner’s survey of 5,728 customers in December 2023, the top concern about AI in customer service was that it would get harder to reach a person, ahead of AI taking jobs and AI giving wrong answers. In its 2026 survey of 3,566 customers, 87% said it’s essential for a company using generative AI to offer a way to reach a human agent. Among customers who weren’t willing to use AI, the most common thing that would change their mind was being able to switch to a person.

This is where a founder has the edge, because you are the person. Put an email address on your contact page, and make sure your chat hands over when someone asks. Refunds, bug reports and cancellations should always reach you, as how to reduce support tickets explains.

That’s the job of human handoff in Deacon. When a visitor asks for a person, or leaves their email after Deacon couldn’t answer, the owners on your team get an email. You reply from the dashboard. If the visitor still has the chat open, your reply appears there under your first name and your company’s, such as “Maya from Acme”. Deacon stops answering while you hold the conversation, and handing it back is one button. If they’ve already gone, they get your reply by email, as long as they gave their address in that conversation.

Whether they have to repeat themselves

Customers hate telling the story twice. In the same Zendesk survey of 6,182 consumers, 74% said they’re frustrated when they have to repeat information, and 81% want agents to continue the conversation without backtracking.

At a small company, the repeating happens at the handover. A chatbot passes on an email address and nothing else, so your first reply has to ask what the problem was. A form asks for account details they already typed into the chat. Read the whole conversation before you reply, and never ask for something the customer has already told you.

Deacon keeps the story in one place. Every conversation is saved word for word, and when you open one you see every turn, the page the visitor asked from and whether they’re new or returning. If your site passes details such as the visitor’s plan, Deacon takes them into account when it answers and shows them beside the conversation. The conversation follows the visitor from page to page and survives a reload, so clicking a link doesn’t make them start again. When a question is too vague to answer, Deacon asks one question back, and after three it hands over with one sentence summing up what the visitor said.

Whether you answer in their language

People would rather ask for help in their own language, even when they read English well. In CSA Research’s survey of 8,709 consumers in 29 countries, published in July 2020, 75% said they’re more likely to buy from the same brand again if customer care is in their language. Among the respondents most confident reading English, 60% still favoured customer care in their own.

A small SaaS that sells in English has customers who read well enough to sign up but struggle to describe a problem in English. Hiring support for every language they speak isn’t an option, and nor is translating every page of your docs.

Deacon replies in the language of the visitor’s latest message, whatever language your docs are written in, so English docs can answer a question asked in Spanish or Japanese. In the dashboard you can translate a whole conversation into your browser’s language. Analytics › Widget lists the languages people wrote in once anyone writes in something other than English. Multilingual chatbot covers how it works.

How to check your own customer service experience

You can check all five in an afternoon, using the messages you already have.

Read your last 30 conversations

Read them as the customer would. For each one, note whether the first reply settled it, whether they had to write again, whether they repeated something they’d already said and what language they wrote in. The share settled by one reply is your first contact resolution. Then note what each one was about, because a how-to that comes up five times is a page to write. Customer insights covers what each kind of question tells you.

Time the waits

Note when each message arrived and when the customer got an answer that helped, in calendar hours. Take the middle wait, then look at the three longest. They’re usually the ones sent at night or at the weekend, which shows you when nobody is there to answer.

Try it as a customer

Email your support address from a personal account on a Friday evening, and see when the reply comes. Then ask your chat something your docs cover, something they don’t, and to speak to a person. Ask once in another language, and reload the page halfway through. Every place you get stuck is a place your customers get stuck too.

Ask customers, if you want a number

A CSAT survey asks customers to rate one conversation, and a customer effort score asks how easy it was to get help. At a few dozen conversations a month, read each answer rather than trusting the average, because two unhappy customers can move it a long way.

Checking it in Deacon

Deacon makes each check quicker. The Playground is the real widget, so you can try it as a customer without using up answers. Every conversation is saved and searchable by what visitors asked. Topics group similar questions, count the people who asked and label most groups as a how-to, a bug report or a feature ask, so the how-tos show you the pages to write and the feature asks show you what customers want next. Once one page’s questions go unanswered more often than anywhere else, Analytics › Widget names that page.

Questions about customer service experience

How is customer service experience different from customer experience?

Customer experience covers everything customers go through with your company, from their first visit to cancelling. Customer service experience is the part where they ask for help. CX meaning sets out the rest.

What does “describe your customer service experience” mean in an interview?

It’s a different question in the same words. An interviewer wants to hear about your own work helping customers, and job sites’ career advice pages cover how to answer it. This page is about what your customers go through when they ask you for help.

Does Deacon translate the widget itself?

No. Deacon’s replies follow the visitor’s language, but the widget’s own words, such as its buttons and the Ask AI box, stay in English. So do three fixed replies no model writes, including the one visitors see when your plan’s answers have run out.

How long does a conversation stay open?

A conversation closes after twenty quiet minutes, and the visitor’s next question starts a new one. If you reply after that, your reply reaches them by email, as long as they left their address in that conversation.

Can Deacon handle refunds or account questions?

No. Deacon takes no actions in other systems, and it always declines questions about a visitor’s own account, invoice or data. It keeps answering their other questions, and the email box it offers lets the visitor reach you about the rest.

Does Deacon ask customers to rate its answers?

No. The chat has no thumbs up or down and no rating on an answer, and Deacon doesn’t run surveys. It gives you the conversations themselves instead, with the questions grouped into topics and counted.