Comment on page
AI Hub
From the AI Hub in your dashboard, you can set up and manage a variety of AI tools, all under one roof. These tools fall into one of three categories:
- AI Assistants: In-product AI tools that make it easier and faster to build and deploy experiences and handle conversations in Inbox.
- AI Actions: AI-powered Flow actions that can be integrated into a workflow to bring AI to your automations.
- Connectors: 3rd party AI tools that you can install to leverage their capabilities in MessageBird.
Some features in the AI hub require you to install an AI connector before you can use them. These features may cost additional funds.
In the Assistiants section of the AI Hub, you can find the following assistants:
- Agent Assistant
- Flows Assistant
- Studio Assistant
- FAQ Assistant
- Audience Assistant
Agent Assistant requires an LLM Connector, such as OpenAI, Azure OpenAI, or Google Vertex AI.
The Agent Assistant uses AI to help your agents work more efficiently as they chat with customers in Inbox.
Once enabled, agents will have access to the following AI features, directly in Inbox:
- Answer suggestion
- Conversation summary
- Rephrase
- FAQ Answers
The answer suggestion AI feature allows agents to click a button during a conversation with a customer, and generate a suggested response, based on the history, topic, and tone of the conversation.
Agents can then choose to send the message as it was generated or customize it to fit their needs, without having to write an entire response from scratch.
In situations where one agent has to hand a conversation over to another agent, or a new agent wants a quick run-down on the conversation so far, they can click a button to generate a conversation summary.
This will create a short internal note that will be added to the conversation, summarizing the exchange between customer and agent so far, and highlighting any key points, such as ‘customer complained about slow delivery’, or ‘agent said they could provide a discount for the customer’.
Sometimes an agent’s message—or an answer suggestion message—doesn’t have quite the right tone. It might be too happy when a customer is airing a serious complaint, or too serious when a customer is leaving good feedback.
The rephrase button allows customers to choose from a variety of tone options, and use AI to re-write their response in that tone.
To use FAQ Answers, you must first set up an FAQ model via the Generate FAQ Answers section in AI Hub.
Agents spend a lot of time handling common questions. The FAQ Answers tool allows them to quickly provide answers to FAQs by automatically generating an answer to a customer’s question from a pre-existing FAQ model and sending it to them directly from Inbox.
To use FAQ answers, you must select the model that you want to use from the drop-down during the Agent Assistant configuration.
The Flows Assistant uses AI to simplify the creation of advanced flows. Once switched on in the AI Hub, the Flows Assistant will be visible for anyone in your organization to use as they build a flow.
All you need to do is describe the flow that you want to build, using a simple written prompt, and the Flows Assistant will generate customizable flow tailored to your needs.
The Studio Assistant uses AI to speed up the creation of message templates in Studio. Once switched on, the Studio Assistant will be visible for anyone in your organization to use as they build a message template.
All you need to do is describe the message template that you want to build, using a simple written prompt, and the Studio Assistant will generate a customizable message template tailored to your needs.
The FAQ (frequently asked questions) Assistant uses AI to automate the creation of FAQ model by extracting information directly from your website or help center.
All you need to do is provide a URL to a website or help center that you want to create, and the FAQ Assistant will generate an FAQ model for you.
From here, you can review the model and edit it as required, before using it to set up an FAQ bot, or as a reference for the FAQ Answers feature in the Agent Assistant.
The Audience Assistant uses AI to make it easier and faster for you to create targeted, granular audiences in Contacts.
Instead of using the audience builder to manually select and define the attributes of the contacts that you want to include, you can type a description of the attributes you want to include, and the Audience Assistant will generate an audience for you. You can then use the audience builder to refine the audience further, as required.
AI Actions are advanced flow actions that you can integrate into your automated workflows. The available actions are as follows:
- Detect intent
- Generate FAQ answers
- Detect languages
- Detect sentiment
- Detect named entities
The detect intent action allows you to use intent recognition, a type of natural language processing (NLP), to understand and label the intent of a customer’s message. Detecting intent is useful for setting up Flows branches, where each branch handles a different type of customer intent.
We also provide a selection of pre-made. pre-trained template intent models for various verticals, including e-commerce, logistics, and retail, that you can use to get up and running with intent recognition faster.
The following languages are supported:
- English (full support)
- French (full support)
- Spanish (full support)
- Portuguese (full support)
- German (partial support)
- Italian (partial support)
- Dutch (partial support)
- Chinese (partial support)
- Arabic (partial support)
The generate FAQ (frequently asked questions) answers action allows you to use FAQ detection, a type of natural language processing (NLP), to understand what a customer is asking and then reply with the specific answer to that question. The generating FAQ answers action is useful for FAQ chatbots that are able to handle and deflect customer questions, or handover to human agents when they’re unable to answer the question.
To use the generate FAQ answers action, you’ll first need to setup an FAQ model. You can then deploy this action within a Flow by deploying the generate FAQ answers action together with messaging, conditional, and loop actions.
The following languages are supported:
- English (full support)
- French (full support)
- Spanish (full support)
- Portuguese (full support)
- German (partial support)
- Italian (partial support)
- Dutch (partial support)
- Chinese (partial support)
- Arabic (partial support)
Build an FAQ answers model
Click here to learn how to build an FAQ answers model.
The detect languages action allows you to use language detection, a type of natural language processing (NLP), to understand and label the language from a customer’s message. The detect language action is useful for automated conversation routing for either human agents or chatbots.
The detect language action comes out-of-the-box so there is no need to configure anything. Simply select the detect language action in Flows, use an incoming message as the source content, and hit publish.
The detect sentiment action allows you to use sentiment detection, a type of natural language processing (NLP), to understand and label the sentiment from a customer’s message. The detect sentiment action is useful for automated conversation routing for either human agents or chatbots.
The detect sentiment action comes out-of-the-box so there is no need to configure anything. Simply select the detect sentiment action in Flows, use an incoming message as the source content, and hit publish.
The following languages are supported:
- English (full support)
- French (full support)
- Spanish (full support)
- Portuguese (full support)
- German (partial support)
- Italian (partial support)
- Dutch (partial support)
- Chinese (partial support)
- Arabic (partial support)
The detect named entities action allows you to use named entity recognition (“NER”), a type of natural language processing (NLP), to understand and label the entities from a customer’s message. The detect named entities action is useful for extracting data from a message for either storage or personalized communication.
The detect named entities action currently supports the following entity types: integer, float, date time, email, and phone number. When setting up this action in Flows, you need to specify the type of entity you want to detect, as well as the locale (e.g. “en”) that you want to support. There is no model configuration needed to use this action.
The following languages are supported:
- English (full support)
- French (full support)
- Spanish (full support)
- Portuguese (full support)
- German (partial support)
- Italian (partial support)
- Dutch (partial support)
- Chinese (partial support)
- Arabic (partial support)
Last modified 1mo ago