Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.dovetail.com/llms.txt

Use this file to discover all available pages before exploring further.

Chat GA Pn

Overview

Have a question about your customer data? Ask Dovetail with Chat. Chat is powered by Anthropic Claude so anyone with workspace access can query customer knowledge in a conversational interface. At the level of a single file, a project, or your whole workspace.

How chat works

Chat auto-applies context from wherever you are in Dovetail, so you’re always asking the right question at the right scope.
  • Focused analysis: Ask about a specific transcript — including OCR’d documents like academic papers — or a single feedback note.
  • Project synthesis: At project level, ask broader questions across all the data in that project: a meta-study, a round of interviews, a collection of sales calls.
  • Workspace overview: Zoom out further and synthesize across projects — different product areas, research themes, or feedback sources.
Chat responds in the same language you write in, and infers sensible search language from your project and document names, so multilingual teams can work naturally. Here are a few ways to put it to work:
Use caseQuestion
Refine your approachAfter reviewing a sales call transcript, ask Chat: “What questions could I have asked to understand the customer’s needs better?”
Uncover gapsWithin a research project, ask: “What adjacent topics are mentioned but not deeply explored in this data?” Use the insights to kickstart recruitment for your next study.
Generate high-level reportsAt the workspace level, ask: “Summarize the most common customer frustrations mentioned across all feedback projects this quarter” for a quick VoC report for leadership.
Accelerate onboardingNew team members can click into different projects and use Chat to ask: “What was the primary goal of this discovery research?” or “Summarize the key recommendations from this project” to get up to speed faster.

Start a new chat

Open Chat from the sidebar, or press ⌘ J (Mac) / Ctrl J (Windows) from anywhere in your workspace. You can expand Chat for more room while you work. Ask anything — from “How might I improve this conversation next time?” on a single customer interaction to “What are the main VoC themes this month?” at workspace level. The real power is conversational: ask follow-ups, push for nuance, challenge a point, or change direction entirely. Treat it like a dialogue with your research, not a search box.

View citations

Every answer is grounded in your data. Clickable citations let you jump straight back to the original quote, source, or document to validate what Chat said.

Turn a reply into a document

When Chat gives you an answer worth keeping, use Create doc from response on that assistant message to turn it into an insight-style document in your workspace — ready to refine like any other doc. Closing prompts (offers to help further, “is this useful?” lines) are stripped out automatically, so the document focuses on the substantive content. Requires a plan that includes creating insights and documents. Your workspace may show an upgrade prompt if this isn’t available.

View chat history

Open Chat, then select Show chat history to see past threads. Conversations are named from what they were about, so you can scan the list quickly.

How chat prioritzes data in Dovetail

When you ask a question, Chat identifies what information it needs, then searches within the context you’re in — a tag, project, channel, or a specific recording or note. If you haven’t set a context, Chat searches across your whole workspace. It combines keyword search, semantic search, and filters until it has enough to answer. You can also steer it in plain language — by project, field, contact, segment, and more. By default, Chat doesn’t weight results by quantitative signals, highlights, tags, or object type. It does use metadata on the items it retrieves — fields, linked contacts, companies — so well-structured data produces better answers. For interviews and sales calls, retrieved content can reflect who said what: customer or interviewee versus interviewer or internal facilitator, when those roles are clear in the transcript.

How to control the data chat uses

You can steer Chat in the conversation itself — ask it to emphasize ARR, frequency, severity, or impact when comparing themes, and it will try to follow that when retrieving and summarizing. Workspace admins (on plans that include it) can add Custom context under workspace settings so the same guidance applies by default across Chat, custom agents, and Ask Dovetail. That’s optional — many teams rely on per-chat instructions only.

How chat uses metadata

Chat reads metadata across your data. The better structured your workspace — fields filled in, contacts assigned, segments used consistently — the better the answers.

Copy and retry

You can copy text from any assistant message. If a send fails, use Retry so you don’t lose your prompt.

Who can see what

Chat only searches and cites content your account can already access in Dovetail. It doesn’t bypass permissions.

Share your feedback

Use the thumbs up and thumbs down in chat to let us know how it’s doing. All feedback is shared with the Dovetail team to help us improve the experience.
Yes, if you hover over the floating Chat button and click the up or down arrows to Hide Chat or Show Chat. Bring it back any time with a quick hover.