What is Assembled's AI Agent for Chat?

Assembled's AI Chat Agent provides everything you need to start automatically resolving chat based support inquiries. 

• Automated AI Answers - using your knowledge base and Assist's Workflows product,, automatically answer customer questions via ultra low latency chat. Everything from simple FAQs to complex self serve flows - automate quickly and easily with Assist. 
• In depth Insights - Understand which chats were truly resolved in a satisfactory way without QA or CSAT scores. Easily drill into the overall metrics and outcomes as well as conversation level detail. 
• Seamless escalations - Assist intelligently detects when conversations should be escalated to your support team, before the end user complains or asks for agent escalation. 

 

AI Chat Agent Configurations

Navigate to Setup > Chat Agent to view Chat configurations. Settings will be previewed in the right hand side of the screen. Some settings like handoff method and wrap-up will need to be saved in order to be previewed.

Once you save your chat settings you can also test these out in the Quality > Chat testing page. 

AI Chat Configurations: Content

On the Content tab, you can set a few key settings: Can determine if you want the homepage functionality, can modify intro messaging, can determine if voice input and attachments are accepted, and you can modify terms and conditions.

Homepage: 

Enabling the Homepage feature expands the chat interface so that end users open up to a homepage  type interface - which gives end users the option to view top resources, or ask a question via chat. 

With Homepage	Without Homepage

 

Terms and Conditions: 

If you enable Terms and conditions, they are visible when hovering on the "Powered by Assembled" tag line. 

 

 

AI Chat Configurations: Styling

On the Themes tab you can modify all the styling, headers, and iconography. 

 

AI Chat Configurations: Escalation and Handoff Method

On the Handoff tab you can configure both the escalation sensitivity and handoff method. 

Handoff Methods:

Live agent	Seamless handoff to a live agent. Requires compatible live chat settings in your contact center.
Open email support ticket	Handoff from chat to email. Your Support team will follow up with end users via email.
Send a link to support	Give end users a link to fill out a support form. Your support team will follow up via email.
No handoff	No ability to escalate to your support team.

 

Live Chat Settings Compatibility: In order to escalate to a live agent in Zendesk, you must be using Zendesk Sunshine Messaging. 

 

Escalation Sensitivity: Assembled AI Agent will intelligently determine when to escalate conversations to your support team based on a variety of factors, including, but not limited to: 

• Negative sentiment: If end user is getting frustrated or is unable to get what they need from chat conversations
• Sensitive subject matter: if end user is highly upset, is trying to contact support about churning, or other major issues. 
• Stated intent: If the end user asks to speak with support team, asks to speak to a manager, etc. 

The escalation sensitivity setting lets you control whether you'd like the AI agent to be standard sensitivity: conversations will be intelligently escalated, but have slightly higher threshold before escalating, or cautious sensitivity: lower threshold before determining to escalate. 

💡	We recommend starting with Standard Sensitivity, and only adjusting to Cautious sensitivity if you deem that necessary during testing phase!

 

Collect Customer Information: 

When a conversation is escalated from the AI agent to your support team you can collect customer information to populate the support ticket. 

By default, we ask for Name and Email, but you can ask for any data that you'd like by setting up Info Fields - and referencing them in the collect customer information section. 

 

AI Agent Configurations: Wrap-Up and Closing Surveys

On the Wrap-up tab you can set up flows for how to wrap up conversations. A wrap up prompts the end user if they have any other questions after a period of inactivity, in order to close out the conversation, or answer any final questions. 

Closing Surveys: After the wrap up is complete you can either close out the conversation or ask for a follow up survey. 

To configure a wrap up flow, set the following key fields: 

• Confirmation message wait time: How long to wait after inactivity before sending the confirmation message to confirm that the user doesn't need help with anything else
• Quick actions: in response to confirmation message users can see two quick buttons in the chat in line. This is an easy way for the end user to quickly reply in line without typing. Users can also type their response in though. 
• Closing message wait time: If user does not reply to confirmation message, how much longer should chat agent wait before sending closing message. 
• You'll also be able to customize the Confirmation message wording, closing message wording, and follow up message wording. 

End user experience - quick actions display during wrap up

 

Closing Survey Example: Closing surveys can be customized to collect CSAT, Customer Experience scores, or other general feedback. In the closing survey set the following fields: 

• Survey introduction message: introduction blurb asking for survey responses
• Question: Name of survey question
• Ratings: Options for survey responses. Note that the Value field is used to aggregate response results, but is not displayed to the end user. 
• Additional Comments label: (Optional) field name for additional comments

Closing surveys will look like the following if configured:

End user experience

 

AI Chat Configurations: Setup

The final settings page for your AI chat agent is the Setup tab. Here you can filter knowledge, add context, and get the installation script instructions. 

Knowledge filtering: If you have multiple sub brands (e.g. you're a company with multiple distinct websites), or have two different chat instances for different audiences (e.g. you're a marketplace company with a dedicated chat for Buyer portal vs Seller portal), you can filter knowledge so that only the relevant articles are displayed in each chat instance. You can filter knowledge to any knowledge tags that we've ingested from your documents. Go to the Knowledge tab to see the list of tags available on your documents. 

Company Context: You can provide additional context or instructions to help improve the chat prompt. We suggest working closely with your implementation manager here. 

Setup: Finally last step is to install the chat widget on the website(s) you'd like it available. Here you can grab the install key. Please see Chat Installation article for additional instructions. 

 

AI Chat Configurations: Profiles

You can create multiple chat instances if you want chat to show up differently in different locations. Use cases: 

• Multiple sub brands - your company has multiple distinct sub brands that end users can visit. Each sub brand / website should have it's own branded chat. 
• Distinct audiences - you have different logins for different audiences. E.g. a marketplace company with a portal login for buyers and sellers - each portal could have it's own dedicated chat that serves up answers for each audience. 
• Distinct contexts - you want to put chat on your marketing site, and your in-product experience - and want to have different styling or messaging for those different contexts. Each of those websites can have a different profile. 

All of these use cases are great reasons to create multiple chat profiles. Note that profiles should map to different install locations - you should not have one webpage with two chat profiles. 

 

To create a new profile: In the header of the chat agent configuration, click the ... menu, then you can duplicate or create a new profile from scratch. 

Once you have multiple chat profiles, you can choose the specific chat profile you want to edit when you navigate to the Chat Agent menu. 

 

Activation

Once you're ready to turn on chat you have a few options: you can activate for all users, or you can gradually roll out and activate for a subset of website visitors (based on webpage cookie). You can also use the advanced installation settings to override activation within your website.