What are generative replies?
Generative replies in AI agents - Advanced is a generative AI-powered customer service agent leveraging Large-Language Models – including those based on a GPT (Generative Pre-Trained Transformer) architecture, the technology behind OpenAI’s ChatGPT.
Generative replies learns from the information in your help center to provide answers to customer support queries in a human-like, conversational manner. Generative replies require no training at all and once connected to your help center can be live and answering customer queries in minutes.
Overall, generative replies aim to enhance customer experiences by providing a responsive and helpful AI agent that can assist customers 24/7 in 100+ languages.
Which channels can I use generative replies in?
In messaging, generative replies can be used to have automated, human-like conversations with your customers based on information in your knowledge base and customer systems.
In email, generative replies can support your agents by generating suggested responses to inquiries in your CRM, which agents can then edit and send.
Do I need a minimum number of support inquiries for generative replies?
No, you do not need a minimum number of tickets. All you need is a comprehensive knowledge base for generative replies to learn from to get started. See this article for some tips for preparing your knowledge base for generative AI.
What languages do generative replies support?
Generative replies are fully multilingual and can converse in 100+ languages. Likewise, it is capable of reading help centers written in any language.
How do generative replies work with dialogues and workflows?
Generative replies provide fully generated responses based on the information in your help center. For some topics, you may want greater control over the actions an AI agent takes. For example:
- If you want the AI agent to execute an action, such as adding tags, routing, updating fields,
- If you want to personalize the response based on customer information from internal systems,
- If you want to use interactive rich media in conversations, such as buttons, forms, and carousels.
For these topics, you would design a scripted dialogue or workflow for the AI agent to follow. All other topics would continue to be responded to using generative replies. In the future, customers will be able to combine generative elements within dialogues, to make the conversations even more natural and spontaneous.
Can generative replies use content from multiple help centers?
Yes, generative replies can be connected to multiple help centers. This can be to centralize knowledge from various areas, or to vary which help center it uses to respond to a customer query based on specific rules, e.g using the North American help center if the customer is based in the USA.
How long does it take to get started with generative replies?
If you are using Zendesk as your help center, generative replies can be up and running in minutes. All it needs is your help center URL. If you are using a different help center, we require 24-48 hours to ensure your help center is successfully imported.
What do I need to get started with generative replies?
You need a public help center (written in any language) for the AI to learn from. The more comprehensive your help center, the better! In minutes, generative replies can connect to your help center using the URL if Zendesk, or imported CSV if other, and instantly begin providing automated, human-like responses to your customer requests.