> ## Documentation Index > Fetch the complete documentation index at: https://docs.vanticlab.com/llms.txt > Use this file to discover all available pages before exploring further. # Using the chat > What the conversational interface can answer, what it cannot, and how it behaves before and after a Friction Score is published. ## What the chat is for The conversational interface turns your **derived** workflow patterns (and optional **Context Hints**) into plain-language explanations, comparisons over time, and suggested next steps. It is grounded in what Vantic has already measured from structure — not in reading your files, screen, or messages. ## Before vs after a Friction Score **Before the confidence gate is met** — You may see **Friction Pulse**-style signals: single factual statements tied to what was directly detected, without a headline Friction Score. That is intentional; a premature composite score would be misleading. **After a Friction Score is published** — You can ask about the score, each dimension (Tool Fragmentation, Workflow Consistency, Context Switching, Workflow Duration), named workflow patterns, and how they move between signal detection windows. ## What it will not do * Read email bodies, documents, spreadsheets, or chat content * See your screen or run OCR * Access other machines or accounts * Change settings or applications without your explicit action ## AI and data flow To generate answers, the product sends **structured, derived pattern context** and your **prompts** to a third-party AI provider. **Raw workflow signals are not sent** for that purpose. Customer content is not used to train those models. Binding detail is in the [Privacy Policy](https://www.vanticlab.com/privacy). ## Live or external facts For topics that change frequently or depend on a third-party site (for example public pricing pages, regulations, or vendor announcements), the product uses [Graceful Web Redirect](/help/glossary#graceful-web-redirect) behaviour: expect a link you can open and verify rather than a potentially outdated answer from the model's training data. ## Where to go next How to read bands, dimensions, and recommendations. Give the chat stable business context without changing the score.