Skip to content
Book a call
en

AI features

At two in the morning the place is closed. The assistant answers for you.

It uses the phrases you taught it and the information on the customer profile. If a person is needed, it hands you the conversation and tells you why.

5 channels
WhatsApp, website chat, Instagram, email and the phone. It stays a single conversation: even if the channel changes you carry on from the same point.
5 reasons
Not every conversation reaches you for the same reason. RelAI tells five of them apart, so you know where to pick up from.
3 states
Active, paused or draft. You prepare it and try it out before it starts answering customers.

The questions come in exactly when you cannot answer

What time do you close, is there room for six on Saturday, can I bring the dog, where do I park. They are always the same questions, but they come in while you are working, from different channels, or when you have already closed.

With RelAI the assistant answers using the replies you gave it. If the conversation goes beyond what it knows, or a person is needed, it hands it to you and tells you the reason.

And it does not start from scratch. When the person writing already has a profile, the assistant finds what you know about them there; the messages you sent them stay in the same thread and, if they ask for a table, the book is the real one. So the answer starts from who you are talking to, not only from the question they wrote.

The same questions, every day
While you are working no reply
From different channels scattered
When you have already closed they wait
and how many walked away?
With an assistant and one inbox
67% closed without you
The rest reach you in full, with the conversation already there and the reason it ended up on your desk.
When they write to you With what you taught it

FAQs, opening hours and ways of asking are set by you. The AI assistant answers as long as the request falls within what it knows.

When a person is needed With the reason already written

The whole conversation arrives in the messages, with what has already been said and the reason a person is needed.

When you are the one asking You decide if it goes

You write the copilot a request in your own words. Then you choose whether to run it now or leave it among the tasks to approve.

It answers. It involves you. It suggests.

When the customer writes, the assistant answers with what it knows and involves you in the cases that need a person. If the request starts from you, the copilot prepares the work and leaves you the final decision.

One job at a time. No improvising.

There is no assistant trying to know everything. You prepare one for bookings, one for information or one for events, each with its own answers and its own hours.

When you step in. You know where to pick up.

The assistant does not have to answer everything. When it stops, it leaves you what you need to understand what happened and carry on from there.

Reason You know it straight away

A complaint, an explicit request, a question it cannot answer. The reason also changes what you have to do next.

Context The conversation is already there

You find everything the customer and the assistant have already said to each other and you do not have to ask them to start over.

Author It is always recognisable

In the conversation log every reply says whether it comes from the assistant or from a person. That way you know who said what and you do not risk two of you answering.

Profile The profile stays alongside

Tags and customer information stay linked to the conversation: before answering you already know who you are talking to.

The five reasons

Five different cases

5 declared reasons
1 inbox for five channels
0 invented answers
It does not know The question goes outside what it knows.
They ask for you The customer wants to talk to a person.
They got angry The conversation turned into a complaint.
It got stuck The replies are not leading anywhere.
You took it over Somebody on the staff steps in directly.

Calls, WhatsApp, Instagram, email and website chat all end up in the messages. From there you see what is still waiting for an answer without having to check five different places.

You tell it what you want. Then you choose what happens.

The copilot starts from what you want to achieve. Then you choose whether to have it run straight away or to look at it before it happens.

The copilot

You ask it the way you would say it

Before it goes

You can leave it waiting for approval

What you want to check ends up in the copilot tasks. From there you look it over and decide whether to go ahead.

Waiting for your yes 3
Approved today 2
Run straight away 4
On your numbers

It suggests what to look at

The guided strategy puts together the data you already have in RelAI and suggests moves worth weighing up. Next to them you find the numbers it starts from: the suggestion does not arrive without context.

Suggestions this week 6
Taken up 2
Left to you 4
The model

You know what is underneath

For the copilot and for writing texts you choose the provider and use your own key from the settings. Where a feature uses rules or keywords instead, it is treated as such: different things are not called AI just because they look clever. And where a figure does not exist it is flagged, not estimated.

1 key, your own
3 different things called AI
0 invented numbers

Thirteen tools. The last word stays yours.

AI assistant One for every job
Scopes What it can talk about
Assistant FAQs The answers you write
Intents How it recognises what they ask
Hours When it answers
Conversation log Who wrote what
Messages Every channel in the same place
Switchboard The calls it answers
Routing Where it sends every call
Call-backs Who has to be called back
Copilot You ask it what to do in words
Copilot tasks What is waiting for your approval
Guided strategy The moves it suggests

Questions about the assistant

What happens if it does not know the answer?

It stops and the conversation passes to you, with the reason. You also find the context up to that point, so you can carry on without making the customer repeat everything.

Do I write the answers myself?

Yes. You start from the FAQs, then you set hours and intents. The real conversations show you new ways customers ask the same thing, so you can add them when you need to.

Can I have more than one assistant?

Yes. You prepare several for different jobs and give each one its own scopes, content and hours. If the subject changes, the conversation can move to the right assistant.

Does it answer the phone too?

Yes. Calls are routed by reason, and the ones left unanswered end up among the call-backs. That way you know which ones to return without starting from the list of missed calls.

Does the copilot do things on its own?

Only if you ask it to. You can have what you asked for run straight away, or leave the task waiting for approval and check it first.

Which model does it use?

It is the right question to ask anyone selling features with AI built in, and the answer has two parts: the model provider and the key, both configured from the settings. Those two parts decide the costs, where the texts written by your customers end up and what happens the day you want to change provider.

Which customer data can it read?

In the configuration you see whether the customer profile is enabled for that assistant: you decide whether it can use it.

How do I tell where a suggestion comes from?

The suggestion shows the counts it is based on. If a figure is not there, it says so instead of filling the gap with an estimate.

Can I try it before turning it on?

Yes. An assistant in draft can be tried without putting it in front of customers. When you are ready you activate it, and you can pause it again whenever you want.

The next question. RelAI answers.

Try RelAI for free or book a call: we prepare the first answers together and put the assistant to the test on one channel.