Train an AI chatbot on your own data

Step by step

Train an AI chatbot on your own data, one step at a time

1. 1

   Create your agent

   Select Agents in the dashboard and create a new agent, or finish onboarding: website, identity, starter questions, and install.

2. 2

   Crawl your website

   Go to Training Data, then Website, and enter the full public address including https. Review the discovered links, submit your selection, and wait for each page to show Indexed.

3. 3

   Upload files

   Open Training Data, then Files, and add PDF, DOCX, PPTX, CSV, XLS, TXT, Markdown, HTML, or JSON files up to 10 MB each. Keep the page open until each upload completes.

   Good to know: Scanned PDFs need searchable text. Image-only pages and files with almost no text are rejected.

4. 4

   Add instant answers

   For questions you want answered word for word, add a Q&A pair: write the question in common visitor wording, format the approved answer, and optionally attach up to five media files.

5. 5

   Set the scope and tone

   Open Widget, then AI rules, choose an instruction preset, and refine it with Edit as custom so the agent knows what to answer and what to decline.

6. 6

   Test before you launch

   Once sources show Indexed, ask questions with known answers, questions in real customer wording, and questions the agent should decline. Fix gaps before installing the widget.