AI agent, chatbot and live chat The same knowledge. A dialogue.

The AI agent in the chat goes through the reviewed knowledge and, when the customer is identified, their orders and deliveries. Then it answers directly. When that is not enough, the conversation is handed to a team member who sees what has already been said.

Short answer

In Supportifier's chat, an AI agent answers customers directly from the reviewed knowledge base and, for identified customers, their orders and deliveries in connected systems. When the material is not enough it says so and hands over to a team member in live chat, who sees the whole conversation. Outside opening hours the contact options are shown instead of a queue, and the question can be sent as a ticket. What is handed over, with which context, is decided during the rollout.

In the conversation

INTERACTIVE EXAMPLE · No question is sent. No customer data is fetched.

In the conversation

The same knowledge. A dialogue.

The chat answers from the reviewed knowledge. When that is not enough, the conversation can be handed to a team member, who sees what has already been said.

A person takes over — without the customer starting again.

Read about AI and live chat The same question in all five views:Can I change my delivery address?
On your websiteEXAMPLE DATA
Your customer serviceAI and live chat · Example conversation

Can I change my delivery address?

Answer from the knowledge base

The address can be changed before the order leaves the warehouse. To check your particular order we first need to identify you and fetch the order status.

The same knowledge sourceDelivery address changeRead the source

Delivery terms: The address can be changed before the order leaves the warehouse.

One reviewed rule is used in all five examples. The public help centre shows published content only. Internal instructions and personal customer data are not public.

Different ways in — not five steps the customer has to pass through.

01

A chatbot that answers from your knowledge.

The answer in the chat comes from the same reviewed articles the help centre shows. That makes the chat answer the way the team would, and lets it say so when the question is outside the material.

Grounded answers
The AI agent answers from what the knowledge base covers. When customers ask about their own order, it fetches order status and tracking from connected systems once the customer is identified, and says so instead of guessing when the detail is missing.
Follow-up questions
The customer can ask a follow-up in the same dialogue. If the answer is that a team member needs to look into it, the chat says so.
Only does what you switched on
For an identified customer the chat can cancel an order that has not shipped, open a return or issue a discount code as compensation, but only the actions you have explicitly switched on. Everything else goes to a person.
02

A person takes over. Without the customer starting over.

When the chat is not enough, the customer can ask for a team member, who takes over the same conversation and sees all of it, including what has been checked and what has not. The customer does not repeat themselves.

During opening hours
The customer is queued and a team member takes over the conversation in the chat. The offer is shown only when live chat is switched on and someone is available.
Outside opening hours
The contact options are shown instead of a queue: send the question as a ticket, read the answer in the help centre or call during opening hours. A ticket lands in the inbox with the whole chat as material.
03

Live chat and AI in the same box.

The chat is one surface, not two. The AI answers first on what it can, and a team member takes over when needed. The same widget on your website does both, and the same knowledge base sits behind it.

04

You set the boundaries.

Which questions the chat may answer, when it should hand over and what should follow to the team member are decided together during the rollout. Automatic handling is used only within the scope and with the controls you have decided on.

05

Common questions

Can the chatbot give wrong answers?

A chatbot that answers freely can make things up. Supportifier's chat answers from the reviewed articles in the knowledge base and, when the customer is identified and a connection exists, the customer's own details. It says so when the material is not enough instead of guessing. You review the content it answers from, and a team member can always take over.

What happens when no team member is available?

The contact options are shown instead of a queue: the customer can send the question as a ticket, read the answer in the help centre or call during opening hours. A ticket lands in the team's inbox with the whole conversation as material. Opening hours and options are decided during the rollout.

Does the team member see what the customer already wrote?

Yes. When a team member takes over the conversation they see all of it, which rule the chat has shown and what has not been checked, for example order status. The customer does not have to start over.

Can the chat retrieve the customer's order status?

Only if a connection to your business system is part of the rollout and the customer is identified. Without that, the chat answers with the general rule and leaves the personal answer to a team member. Reading order status and changing an order are also two different permissions.

Do we need an English and a Swedish chat?

No. The chat, help centre and form are available in Swedish, English, Norwegian, Danish, Finnish and German, and the widget follows the language of the page it sits on. Answers come from the knowledge base, so which articles should exist in more languages is something we go through during the rollout.

Next step

Start from your everyday work.

Tell us which question takes time today. We go through what a first step could look like.

A clear first step beats a big promise.

See how the knowledge work, the review and the first channel fit together.

How the rollout works
From the same question
to a better answer.
Book a walkthrough