Product
Everything you can hand overSee every feature →
Use cases
What Felesh looks like for youAll use cases →
PricingTrust
Resources
Log in

How do I answer customers' messages at night without a night shift?

An after-hours chatbot should answer from what you taught it and hand the rest to you with the reason. Here is how that works on Felesh.

5 min read · · The Felesh team

Why night messages pile up

Customers do not keep office hours. They write while dinner is cooking, after the children are in bed, or in the quiet hours before dawn. The message arrives on the website chat or in the Instagram inbox, and the phone stays dark. By morning there is a pile of questions: can you hold this order, what time do you open tomorrow, is the price the same as last week.

A late answer can read as not caring, even when you were simply off for the evening. The shop is closed, nobody is sitting by a phone, and nobody wants a night shift. The usual response is to carry the phone around and hope. A better one is to decide, in daylight, what a night-time reply should be.

The short answer: teach Felesh your answers during the day; at night it answers from them, and what it cannot answer waits for you with the reason.

Fig. 1 Night messages are answered from what you taught, and what it cannot answer waits for you with the reason.

What the bot should do after hours

A chatbot that works after hours has a short job description. It answers the questions it can answer, and it says plainly when it cannot.

People who write at night usually want the same few things: a confirmation, an opening time, a price, an honest “we will look at this”. The chatbot’s main job is to carry those answers well. It also has to know the edge of its own knowledge. A question about your returns policy is fine if the answer is in the material it was given. A question about an order that went wrong in a way nobody has described is not.

The bot should not guess, because a confident wrong answer is worse than none. A person behind a counter does the same: answers the easy question, calls someone over for the hard one, and leaves a note when the counter is empty.

So the test of a bot that works after hours is not how much it can say. It is what it does at the edge of what it knows.

Teach it before you leave

The teaching happens during the day, when the shop is quiet and you can concentrate. Felesh answers from what you have taught it. You teach it by uploading documents and writing instructions, and it answers from those.

Start with the documents you already keep: your delivery policy, your returns policy, your price list. Write the instructions in the words you would use across the counter. If a customer asks about returns, you want the answer in your own words, not a stranger’s.

Cover the questions that stop someone placing an order: postage, timing, cancellation, faulty goods. If you close on certain days, say so in the instructions. A customer at night is not after an essay. They want a clear answer, and that is what the instructions are for. Keep them short, in plain English, and in the order a customer would ask.

Do not try to teach it the whole history of the business. Teach it the questions that come up every week. When the business changes, change the instructions with it, because the answers are only as current as the last update.

Where the messages arrive

Customer messages from your website chat, Instagram, WhatsApp and Telegram arrive in one inbox. You do not need a separate routine for each channel, and you do not need to check two places in the morning.

That matters for a small team. One place to look means one place to teach, and one list to work through when you start the day.

What happens to the questions it cannot answer

No bot can answer everything, and the honest ones do not pretend otherwise. When the material you taught Felesh does not cover a question, it does not make an answer up. The message goes to the “Needs you” list, with the reason it could not answer.

That phrase “with the reason” matters. When you open the list in the morning, each waiting message says why it is there, so you can tell a simple request from a messy one. You decide which is urgent and which can wait.

Nobody is notified. Nothing wakes you at midnight to say a message has arrived; the case simply waits until you look. And when a person joins a conversation, Felesh stays quiet, so you never find it talking over you.

The morning routine is short. Open the inbox, read the reasons, answer what needs you. That is a fair exchange for not working a night shift.

Before a holiday rush

Before a holiday or a busy week, set aside time to update what you have taught it. Customers will ask about opening hours, dispatch deadlines and returns periods. Add those answers to the instructions before the rush, not during it.

Whatever the instructions do not cover lands in Needs you, each with its reason, and you will see which gaps to fill next.

That is a useful side effect. The “Needs you” list is a record of what customers ask that you have not yet written down.

Try it on one channel

If you are unsure, start with the website chat. Keep the trial small: teach it your returns policy, your opening hours and your delivery times. Look at the “Needs you” list when the morning comes, and see what it could not answer.

The point is to see it for yourself. See how the inbox works, and what hand-off shows when it cannot answer.

Next question Do I have to tell customers they are talking to an AI? What Article 50 of the EU AI Act says about telling people they are dealing with an AI, and who the article addresses.