The answers provided have a strict confidence score when using the auto-reply mode. If the confidence score falls below a certain threshold, the system will escalate the query for you and your team to handle. In draft mode, the confidence score threshold is lower, allowing more leniency as it usually involves a human review before sending the response.

No. Our system will only access the specific email threads or messages necessary to generate a response. Your personal or sensitive information is never shared with us or any other parties. We do not store anything besides encrypted generated responses, nor do we analyze or use any data outside of the email interactions needed for responses.

No human eye reads your emails—everything is encrypted. We follow strict security measures to protect your data. Your emails are end-to-end encrypted, and we do not store or share your email contents. Only you have access to your emails and the responses generated.

All processed content from your integrations is stored in isolated containers. All data is encrypted at rest (AES-256) and in transit. Your data is never used for any reason (including AI model training) other than servicing API calls or customer support at your request only.

An AI agent is an artificial intelligence system that can understand natural language conversations and questions, and provide helpful responses based on its training. InboxPilot is an AI agent platform that allows you to build your own personalized AI agent by uploading your own documents, content, manuals, etc. The AI agent then learns from this uploaded content and can answer questions instantly based on the information provided to it.

For all the amazing things ChatGPT can do, it can't access your company's private content or data. So it can't answer question on how your business or products actually work. With InboxPilot, you get the power of ChatGPT, but with an AI agent that knows everything about your company and it's internal or external documentation.

Around 80% of customer support conversations can be resolved by one of our AI agents. But when a question can't confidently be answered by the AI agent, then your customers are given the option to 'Talk to a person'. If they select this option, they are seamlessly transferred to one of your agents within your existing support tool, e.g., Gmail or Zendesk.

Never.

The chatbot will continue to respond as long as the conversation remains relevant and productive. However, it may stop replying if it detects that the conversation has reached a natural conclusion, if the query requires human intervention, or if it encounters topics beyond its capabilities. Additionally, if you decide to jump in and reply to the email yourself, the chatbot will stop replying to avoid duplicate responses.

No, it's entirely based on the plan you choose. Each plan gets a number of emails allocated per month, and they reset on the first of every month.

When you send an email, our LLM analyzes the content of your message and generates a response based on the context and information it has learned. The model ensures that each reply is tailored to your specific inquiry, offering accurate and helpful answers.

Yes, the email chatbot can process and respond in several languages. If you send an email in a different language, the system will respond using the same language the incoming email was received. For specialized language support, a human representative will assist as needed.

Yes, the email chatbot keeps track of the context of the email thread and answers based on the previously asked questions and relevant data. This allows it to maintain a coherent and meaningful conversation throughout the email exchange.

The system may flag an email for human intervention when it detects sensitive or complex issues, unclear or ambiguous requests, negative sentiment, or specialized inquiries that require specific knowledge beyond the LLM’s capabilities. It may also escalate emails with unusual language, formatting, or technical jargon. Additionally, the system flags a thread for human intervention if the recipient asks to speak to someone, if no relevant answer could be provided, or if the answer provided falls below a confidence score. In these cases, a human representative will review the email and ensure your issue is properly addressed.

To connect your website's form to InboxPilot, you need to link the email account where your form submissions are sent (e.g., Gmail or Outlook) and specify keywords to identify them. In your InboxPilot dashboard, go to the "Connect Email" section and select your email provider. Authorize InboxPilot to access the account receiving your form submissions. Then, in the setup form, under "Reply-Only Keywords (Optional)," enter unique words or phrases from your form emails (e.g., "Contact Form," "Submission ID," or "Support Request") that InboxPilot should look for to process new submissions. Once connected, InboxPilot will monitor that inbox and respond to emails containing those keywords.