AI-assisted live chat
Last updated
Last updated
Activechat offers you a unique combination of a smart natural language chatbot platform with an advanced AI-assisted live chat engine. This allows you to achieve a bunch of things that are quite challenging to do with other similar tools:
Accept incoming chats from your customers or website visitors, route them to the most appropriate team or agent, and use previous similar conversations to get response suggestions.
Build complex automated conversation scenarios that can be triggered by specific natural language intents or keywords or by external events (like adding an item to a shopping cart or visiting a specific page on your website)
Analyze conversational data to understand your customers better and even measure their sentiment in a conversation.
By default, all incoming conversations are handled by your virtual assistant. The goal is to understand what your customer is asking or inquiring about, and provide an automated response or escalate the conversation to a human when necessary.
Some businesses prefer to handle all conversations manually. While we encourage you to use automation as much as possible, this manual approach is also possible with Activechat. Read more about setting up this behavior in our guide to the "default" bot skill which is triggered on every message your customers send.
Live chat sessions can be triggered in multiple different ways:
from the visual builder through the LIVE CHAT block
from a natural language intent like "Talk to a human" or "Agent please"
by the human agent from a CRM interface
When you create a new project, you will find a "Live chat" system intent in the "Automation" menu of the platform. You can customize this intent by adding more phrases that should trigger it, and when your customers say something similar, a new live chat session will be started.
Once the live chat is initiated (either by the agent clicking the "Start chat" button in the CRM or from any of your chatbot skills), it is assigned to a live chat agent based on the assignment rules (group or tag). Agents can see assigned chats in their “Conversations – Live chats” panel.
New chats are assigned only to the agents who have their “Accepting new chats” status set to “Yes”. If this is set to “No”, the agent is still able to communicate in the chats there were previously assigned or re-open chats that were closed, but will not receive new chats (useful for situations when there is a lot of incoming requests and agent wants to concentrate on servicing the ones that are already open).
Activechat is a new generation live chat automation platform that constantly learns from conversations that you do manually to help you automate similar conversations in the future. Every time you (or your support agent) close a live chat session, we use that session data to improve the machine learning model that powers AI hints. These hints can be used by other agents in your organization to improve their performance and provide better support quality.
Every time your customer sends you a message, the platform analyzes previous conversations for similarity and generates a response suggestion. You can click that response to send it to the customer immediately (perhaps, after some editing).
To access AI hints, pull up the "Show AI hints" tab above the message window. If you're not satisfied with the current suggestion, just hit the refresh icon on the right, and the system will generate another one. Click a hint to paste its text to the message window and send it to your customer.
For details on setting the context for these hints and fine-tuning the AI model, please refer to the next section of this manual.
Chats can be transferred to another agent or group by clicking the “Transfer to…” button below the chat and choosing an appropriate agent (or group) from the list.
Agents can transfer chats to other agents who are online and in the "Accepting chats" status.
Once the live chat session is done, it’s important to close the live chat session and resume the bot. The bot can be resumed from any skill. We advise to pre-build a skill that agents can use to ask customers for feedback about the service quality.
Choose a skill to run when the chat session is over.
In case no skill is chosen, the bot will get back to the “idle” mode and the “default” skill will be triggered on the next message sent by the user.
