How to Automate Customer Enquiries with AI (Without Losing the Personal Touch)

Short answer: you automate customer enquiries with AI by collecting the questions your customers really ask, writing approved answers to the most common ones, connecting those answers to a single channel (website chat, email or messaging), and setting clear rules for when a person takes over. Start small, test with real questions, and tell customers when they are talking to an AI.

About this guide

This is a research-based guide. We have not hands-on tested any specific product for this article, so we do not recommend one. Last reviewed: 16 November 2026. See our Editorial Policy for how we label our articles.

Key takeaways

  • Automate the repetitive, low-risk questions first, such as opening hours, pricing basics, delivery and booking details.
  • Keep complaints, refunds, negotiations and anything sensitive with a person.
  • Pick one channel, get it working, then add another.
  • Your approved answers matter more than the tool you choose.
  • Tell people when they are talking to an AI. In the EU this is now a legal requirement for chatbots.
  • Measure a few simple things every week and keep correcting the answers.

What is in this guide

What it means to automate customer enquiries

Most small businesses answer the same handful of questions every week: when are you open, how much does this cost, where is my order, can I change my booking. Automating enquiries means letting software handle those repetitive questions so you and your team can spend time on the conversations that need judgement.

It helps to think of three levels, because they carry different amounts of risk:

  1. Fixed replies. An auto-reply or a set of saved answers that goes out when a message matches a simple rule. Low risk, limited flexibility.
  2. AI drafts, a person approves. An AI tool writes a suggested reply and you or a team member checks it before it is sent. A good first step if you are cautious.
  3. AI answers directly. The AI replies to the customer on its own, using your approved information, and hands over to a person when it cannot help. The most time-saving, and the one that needs the most care.

Many articles on this topic come from software vendors, and they often quote large figures for how much work AI can take off your plate. Treat any such number as a vendor claim until you see how it was measured. The only figure that matters is the one from your own inbox, which you can measure after a few weeks.

Which enquiries to automate first, and which to keep human

Look at the last few weeks of messages and sort them. Questions with a clear, stable answer are good candidates. Questions where a customer is upset, where money or safety is at stake, or where the answer depends on judgement should stay with a person.

Type of enquiry Automate? Why
Opening hours, location, contact details Yes The answer rarely changes and a wrong answer is easy to fix.
Pricing basics and what is included Yes, with care Fine for published prices. Quotes that depend on the job should go to a person.
Order status and delivery information Yes, if connected to real data Only works well when the AI can see the real order status.
Booking changes and availability Yes, if connected to your calendar Needs access to live availability, otherwise it will guess.
Returns and refunds policy questions Yes for the policy, no for decisions The AI can explain the policy. A person should approve exceptions.
Complaints and upset customers No People need to feel heard. Hand over quickly.
Custom quotes, negotiations, large orders No These need judgement and are worth your personal attention.
Anything involving health, legal or safety advice No The risk of a wrong answer is too high.

Start with one channel

You do not need AI on your website, email, WhatsApp and social media at once. Go where most of your enquiries already arrive and start there. If most messages come by email, begin with email. If customers mostly message you on social media, begin there. Each channel has its own quirks, and getting one right first saves you from fixing the same mistakes four times.

Step-by-step: how to set up AI to handle enquiries

  1. Collect your real enquiries. Export or copy the last 50 to 100 messages from your chosen channel. Remove names and personal details before you share them with any tool.
  2. Group them and find the top five. Sort the messages into themes and count them. The five most common questions are your starting point.
  3. Write the approved answers. For each question, write the answer you want customers to receive, in your own voice. This becomes the source of truth the AI relies on. Include prices, policies and links. Keep it short and accurate.
  4. Choose an approach. Pick one of the three setups in the next section based on how much control and how much time you have.
  5. Set the handover rules. Decide in advance when a person takes over: a complaint, a request for a refund, a question the AI cannot answer, or a customer asking for a human. Make it easy for the customer to reach a person.
  6. Tell customers it is an AI. Add a clear line, for example “You are chatting with our AI assistant. A team member can step in at any time.”
  7. Test it with real questions. Ask it the questions from step 1, including awkward ones and ones it should not answer. Check that it hands over correctly and never invents prices, policies or promises.
  8. Launch to a small slice. Turn it on for part of your traffic or only during quiet hours first, then widen it once it behaves.
  9. Review every week. Read a sample of conversations, fix wrong or missing answers, and add new questions as they appear. This is what makes it improve.

Three ways to set it up

Approach Best for Effort Watch out for
Built-in features of a tool you already use (for example an email, helpdesk or booking tool) Businesses that already use one tool for enquiries Low Features and limits vary by plan, so check what your plan includes.
A website chatbot trained on your own content Businesses with a website that gets regular visitors Medium Check how it uses your content, what it does when it does not know, and where customer data is stored.
A general AI assistant that drafts replies you approve Owners who want control and a lower risk start Low to medium Saves writing time but still needs a person to send each reply.

If you want to connect your inbox, forms or booking system to an AI, a workflow automation tool can pass messages between them. That adds power, and it also adds more places for something to go wrong, so build it up after the basic version works.

Whichever approach you choose, ask the vendor these questions before you commit:

  • What happens when the AI does not know the answer?
  • Can a person take over mid-conversation, and how quickly?
  • Where is customer data stored, and is it used to train the vendor’s models?
  • Can I see and edit the answers it is using?
  • Can I leave, and take my data with me?
  • What does it cost when volume goes up?

Pros and cons

Pros

  • Customers get an answer outside working hours.
  • Replies are consistent, because they come from your approved answers.
  • It frees time for work that needs a person.
  • Enquiries that would otherwise be missed get a first response.

Cons

  • A wrong or invented answer can damage trust, so setup and testing take time.
  • Some customers dislike talking to a bot, especially when they are upset.
  • You need to keep the answers up to date as prices and policies change.
  • Customer data is shared with another company, which brings privacy duties.

What it costs

Prices change often and differ by plan, so we are not quoting figures here. Instead, here is what usually drives the cost:

  • The number of conversations or messages you handle each month.
  • The number of channels you connect.
  • Extras such as connections to your calendar, orders or CRM, and extra team seats.
  • Your time to set it up and review it each week. Do not leave this out of the calculation.

A simple way to judge whether it is worth it: estimate the hours you spend each month on routine replies, multiply by what your time is worth, then compare that with the monthly subscription plus the hour or so you will spend reviewing it each week. Also count the enquiries that currently get answered late or not at all, because that is often where the real value is.

Legal and trust basics

Tell people when it is an AI. Being open about it is good practice everywhere. In the EU, rules on AI transparency say that people must be told when they are interacting with an AI system, unless that is obvious. Reports indicate these chatbot disclosure rules have applied since 2 August 2026, and they can apply to businesses outside the EU whose chatbots serve EU customers. Dates and details for other parts of the law have been changing, so check the current text or ask a professional before you rely on a particular deadline.

Protect personal data. Customer messages often contain names, emails, order numbers and sometimes more. If you serve customers in the UK or EU, data protection law applies to how you collect, use and store that information. Choose vendors that explain where data is stored and how it is used, update your Privacy Policy to say you use AI to handle enquiries, and avoid pasting sensitive details into tools you have not checked.

Do not let the AI make promises. Your approved answers should say clearly what the AI can and cannot offer. Set it up so that refunds, discounts and exceptions always go to a person.

This is general information, not legal advice. If you handle sensitive data, or you are unsure about your obligations, speak to a qualified adviser.

Common mistakes to avoid

  • Automating everything at once. Start with five questions and grow.
  • Letting the AI guess. If it does not know, it should say so and hand over.
  • Hiding the human option. Customers should be able to reach a person easily.
  • Skipping the weekly review. Answers go out of date and gaps only show up when you read real conversations.
  • Trusting vendor numbers. Measure your own results.
  • Ignoring privacy. Check where customer data goes before you connect anything.

How to measure whether it is working

Track a handful of simple measures from the start, so you have something to compare against:

  • Time to first reply before and after.
  • The share of enquiries resolved without a person.
  • How often customers ask for a human, which shows where the AI falls short.
  • Wrong or unhelpful answers you find in your weekly review.
  • Customer feedback, even a simple “Was this helpful?” button.

Review these after a month. If the share of enquiries resolved without a person is low and the number of handover requests is high, the fix is usually better answers, not a different tool.

Frequently asked questions

Can AI handle all my customer enquiries?

No, and it should not try to. AI works best on predictable, low-risk questions. Complaints, refunds, negotiations and sensitive topics should go to a person.

Do I have to tell customers they are talking to an AI?

You should. It builds trust, and in the EU the rules say people must be told when they interact with an AI system, unless it is obvious. Check the current rules for the places you serve.

Is it safe to share customer messages with an AI tool?

It depends on the tool. Check where it stores data, whether it uses your data to train its models, and whether it meets the data protection rules that apply to you. Remove personal details from any messages you use for testing.

How long does it take to set up?

A basic version covering your top five questions can take a few hours to a few days, depending on the tool. Plan for ongoing weekly review time as well, because that is what keeps it accurate.

The bottom line

Automating customer enquiries works when you treat it as a small, careful process, not a switch you flip. Start with the five questions your customers ask most, write the answers yourself, choose one channel, make handing over to a person easy, and be open that it is an AI. Review it every week and it will get better. For more guides like this, see our Business Automation section.

How we researched this guide

We reviewed what is already published on this topic, including guidance from software vendors and small-business publishers, and checked the legal points against the sources below. We have not hands-on tested specific products for this guide, and nobody paid for any mention here. Several of the pages we read are written by vendors, so we have not repeated their results figures. If you find a mistake, tell us through our Contact page and we will correct it.

Sources

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *