> ## 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.

# Crafting context in Dovetail

## Overview

Dovetail AI already searches your data for every question. What it doesn't know by default is your company, why a given project exists, or how a channel's incoming feedback should be read. You can give it that background yourself, at three levels — workspace, project, and channel — and each one shapes answers whenever someone chats in that scope.

**Rule of thumb:** workspace context is *who you are*. Project and channel context are *what this work is about*.

| Level         | Where you set it                          | Best for                                                               |
| :------------ | :---------------------------------------- | :--------------------------------------------------------------------- |
| **Workspace** | Linked context Docs in Workspace settings | Company-wide knowledge — product, terminology, voice, strategy         |
| **Project**   | The project overview                      | Why this research exists — goals, methods, scope                       |
| **Channel**   | The channel's context field               | How continuous feedback should be read — product area, audience, focus |

***

## Workspace context

[Workspace context docs](https://docs.dovetail.com/help/dovetail-ai/workspace-context-docs) let admins link existing Docs as persistent AI context for the entire workspace. Once linked, that content is injected into every Chat, Agent, and Ask Dovetail conversation — company overview, glossary, brand voice, customer segments, reporting standards, anything that should hold true everywhere.

Prefer a handful of focused Docs over one long file, share them with the workspace at **Can view** access or higher, and keep them current — stale strategy docs lead to stale AI answers.

<Tip>
  For templates and a full setup guide, see [Workspace context docs](https://docs.dovetail.com/help/dovetail-ai/workspace-context-docs).
</Tip>

***

## Project context

A project's overview is loaded automatically whenever Chat is opened on that project. It's the right place for background specific to *this* study — not the whole company.

A well-crafted overview covers:

1. **Purpose & goals** — research questions, business context, expected outcomes
2. **Terminology** — product names, personas, and acronyms specific to this work
3. **Data sources** — interview types, time period, geography, tools used
4. **Methodology** — how data was collected, frameworks, key stakeholders
5. **Focus areas** — themes to prioritize, and what's out of scope

> *Usability study of the checkout redesign (Jan–Mar 2026). Twelve remote interviews with enterprise buyers who completed a purchase in the last 90 days. Goal: identify drop-off causes in steps 2–3. Out of scope: mobile app and billing. We say "checkout flow," not "cart."*

Put company-wide terms in workspace context and study-specific terms here, and update the overview whenever scope or research questions change.

<Tip>
  See [Crafting your project overview](https://docs.dovetail.com/help/chat/prompt-guidance/Prompt-guidance#crafting-your-project-overview) for the full breakdown.
</Tip>

***

## Channel context

Channels track continuous, high-volume feedback — support tickets, sales calls, app reviews — so the AI needs a standing brief on what a given stream is and what to focus on. That brief lives in the channel's context field, which is loaded whenever Chat is used on that channel.

Keep it under 400 characters, written as natural language rather than a list, and anchored to 2–3 goals rather than an exhaustive spec. State your role so the AI analyzes data from your perspective, and avoid getting so specific that you filter out other valid insights.

> *I'm a Product Manager tracking the Analytics dashboard channel, mostly mid-market admins. Prioritize reporting accuracy, export failures, and permission issues. Ignore billing and SSO — those belong to other channels.*

<Tip>
  See [Context and topic descriptions](https://docs.dovetail.com/help/channels/context-and-topic-descriptions) for the full principles, strategies for multi-team channels, and how to pair context with topic names.
</Tip>

***

## How the three levels work together

When someone asks a question, Dovetail combines:

1. **Workspace context** — always-on background
2. **Location context** — the project overview or channel context, depending on where Chat is open
3. **Chat results** — live evidence from that scope

| You want the AI to know…           | Put it in…                         |
| :--------------------------------- | :--------------------------------- |
| Our product name and brand voice   | Workspace context docs             |
| This study's research questions    | Project overview                   |
| What this support channel is about | Channel context field              |
| What customers said last week      | Nothing to craft — Chat handles it |

***

## Quick-start checklist

* [ ] Workspace: link 3–5 focused context Docs (product, glossary, voice, customers, standards)
* [ ] Each active project: fill in a clear overview — purpose, scope, methods, key terms
* [ ] Each channel: write a short context brief — role, focus, and what's out of scope
* [ ] Test with one question at each level that only background knowledge could answer well

***

Related articles

* [**Workspace context docs**](https://docs.dovetail.com/help/dovetail-ai/workspace-context-docs)
* [**Prompt guidance**](https://docs.dovetail.com/help/chat/prompt-guidance/Prompt-guidance)
* [**Chat: Technical overview**](https://docs.dovetail.com/help/chat/technical-overview)
* [**Context and topic descriptions**](https://docs.dovetail.com/help/channels/context-and-topic-descriptions)