An AI chatbot for customer service, trained on your content
Answer routine questions from your website and help center. Give customers useful source links and bring in your team when a conversation needs a person.
![](data:image/svg+xml,%3Csvg xmlns="http://www.w3.org/2000/svg" width="1" height="1"/%3E)
How it works
Start with the knowledge you already have
Import your website content and help articles, then review the sources the agent can use. Test common customer questions before enabling replies. Update the underlying content as your products, policies and services change.
Preview
Hand over to the team
Language
Answers with a source to read
The AI agent uses your approved knowledge to respond and can link customers to relevant help articles.
AI usage
See the cost of each AI reply in your usage ledger.
Train your AI chatbot on your content
Write answers for specific questions
Add policies, product facts and answers that are not on your website.
Use your help center
Make your published help articles available to the agent, so customers can get answers from the content your team maintains.
Import your website
Add a public website URL to bring its content into your AI knowledge library.
Set the tone and handoff rules
Choose a personality, add instructions and decide when the agent should bring in your team.
Add internal documents
Upload Text or Markdown files up to 1 MB.
Identity verification
Verify the customer before sharing account details
Ask for a one-time email code before your AI uses a protected tool.
Connected tools
Connect your AI to your business tools
Connect an MCP server and choose the tools your agent can use.
Choose the right balance for your AI agent
Explore the balance between speed, capability and cost.
How it works
Website
Preview
Publish
Turn a repeated question into a reviewed answer
Imagine that your team answers the same return-policy question several times a day. Start by checking the article that explains the policy. Import it as an approved source, ask the AI agent several versions of the question, and review whether the answer includes the conditions a customer needs to understand. A useful AI chatbot for customer service begins with clear information and a realistic test, before you enable it for visitors.
After launch, read the conversations that reach a person. A missing detail may call for a better article, while an account-specific exception may need your team. Keep those two cases separate when improving your content. The AI customer service agent and the shared inbox work together: automation handles the questions it can answer from your sources, and a teammate can continue the conversation with its existing history.
Frequently asked questions
Frequently asked questions
What can I use to train the customer service chatbot?
Use your website and help-center content as sources. Review the imported knowledge and test responses before making the agent available to customers.
Can customers speak to a human?
Can I test the AI chatbot before publishing?
Can the AI agent answer in different languages?
Does the customer need to open a new chat?
Which plan do I need?