> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aseed.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Research Assistant
> Use the Research Assistant to generate research context, interview guides, ask questions about project reports with citations, and get help on Aseed.
## Overview
The **Research Assistant** is the AI-powered section in the sidebar (link **Research Assistant** or path `/space/agent`). It helps you prepare and use research context, interview guides, and project documents in one place.
**What you can do:**
Generate a structured Research Context Brief through an interactive Q\&A. The brief defines goals, hypotheses, and method. Save it to a project's Research Goals.
Create or refine interview guides with the AI. Save guides directly to a project or use them in the Interview Guide Builder.
Ask questions about your project reports. Get answers grounded in project documents with citations. Optionally scope to the whole project or to specific reports and records.
Ask questions about Aseed features, workflows, and best practices. Answers are based on up-to-date documentation.
Research context (goals and interview guide) tells Aseed what you want to learn and how interviews should run. You can add it in two ways:
* **AI Generation**: Use the Research Assistant to generate a Research Context Brief through an interactive Q\&A session
* **Manual Entry**: Write or paste research goals and interview guide in the project (Research Goals and Interview Guide cards)
## Ask Project
Ask Project lets you ask questions about your project's reports and get answers that cite the actual project documents.
**How it works:**
1. **Turn on Ask Project** — Click the **Ask Project** suggestion in the empty state or use the **`+`** button and choose **Ask Project**. Select one or more projects. The first time you use a project, Aseed indexes its documents (project report, single reports, transcripts); indexing may take a short time.
2. **Ask a question** — Type your question in the chat. The assistant searches the selected project's documents and returns an answer with **citations** to the source report or record.
3. **Scope to specific documents (optional)** — By default, answers use all indexed documents in the selected project(s). You can narrow the scope: open the project selector and choose specific **reports** or **records**. Questions then use only those documents, so you can "talk" to a subset (e.g. one report or a few records) instead of the whole project.
**When to use Ask Project:**
* Summarize or compare findings across several reports
* Find where a theme or quote appears in the project
* Ask follow-up questions about a single report or a few records
* Check claims against the actual project content
Answers are grounded in your project data; citations link back to the source so you can verify and reuse the information.
## Generate research context with AI
The Research Assistant can build a Research Context Brief by asking focused questions about your project.
The brief includes 11 sections: metadata, decision framing, product context, goals, questions, hypotheses, participants, method, analysis, outputs, glossary, and limitations.
Navigate to the Research Assistant (sidebar link or `/space/agent`). You'll see an empty state with suggestion buttons.
You can start in two ways:
* **Click the suggestion button**: Click **"Create Research Context"** in the empty state
* **Type a message**: Ask the assistant to create a research context brief. For example:
* "Create a research context brief for my project"
* "Help me define research goals and hypotheses"
* "I need to prepare a research context"
You have two options:
**Option 1: Answer questions interactively**
The AI will ask you questions about:
* Research type and stage
* Decision framing (what decisions will be made)
* Product or service context
* Research goals and questions
* Target participants and segments
* Research method and protocol
* Hypotheses to test
* Analysis approach
* Expected outputs
Answer each question so the AI can build a complete brief.
**Option 2: Provide materials**
If you already have notes or materials, share them in your first message (or later when asked). The AI will extract what it can and ask follow-up questions only for missing details.
**Tips for answering questions:**
* **Choose your reasoning mode**: Switch between **Fast** (2 tokens) and **Thinking** (5 tokens). Use Fast for most tasks and Thinking for complex cases.
* **Skip questions if needed**: If you don't know the answer, click the **Skip** button. The AI will ask alternative questions.
* **Long messages are automatically collapsed**: If your message is longer than 500 characters, it will be collapsed with a "Show more" button. Click to expand and see the full message.
After gathering enough information, the assistant will show a summary. Review it and confirm to generate the research context brief.
Once generated, you'll see:
1. **Preview card in chat** — A card showing the first few lines of the brief with a **View** button
2. **Full brief panel** — Click **View** to open the complete brief in a panel on the right side
Use the **Save** button in the brief panel to:
* **Download as Markdown** — Save the brief as a `.md` file
* **Copy to clipboard** — Copy the brief text for pasting elsewhere
* **Save directly to a project** — Select a project from the dropdown to attach the brief to that project's Research Goals field
When you save the brief to a project, it populates the Research Goals field in that project. If the project already has research goals, you'll be asked to confirm overwriting them.
Projects with research context show a **"Has context"** badge. Projects without it show **"No context"**.
## Token usage and cost
The Research Assistant uses tokens for each assistant message when generating a research context brief. You can choose between two reasoning modes:
**Cost:** 2 tokens per assistant message
Use Fast mode for quick responses and standard brief generation. This mode provides fast, efficient answers suitable for most use cases.
**Cost:** 5 tokens per assistant message
Use Thinking mode for more complex research scenarios that require deeper analysis and reasoning. This mode provides more thorough responses and better handles nuanced requirements.
You can switch between Fast and Thinking at any time with the mode selector. Use Fast for speed and Thinking for deeper analysis. Tokens are charged only for assistant messages.
**Example:** If the AI asks 10 questions and you answer each one, that's 10 assistant responses. In Fast mode, this costs 20 tokens (10 × 2). In Thinking mode, this costs 50 tokens (10 × 5). Your questions don't cost tokens.
For more details about token usage and costs, see [Token usage](/account/usage).
## How to add research context manually
Inside your project, scroll to the section below the project header that contains **Research Goals** and **Interview Guide**.
Click on the **Research Goals** card. In the modal, describe:
* What you're trying to learn
* Questions you want to answer
* Hypotheses you want to test
* Product or service context
Click **Save** to store your goals.
Click on the **Interview Guide** card. You can:
* Paste your interview script or topic list
* Use a guide created with the [Interview Guide Builder](/interviews/guide-builder) - you can save AI-generated guides directly to projects
* Write the guide manually
Click **Save** to store your guide.
If you generated an interview guide using the AI Agent, you can save it directly to your project from the guide panel using the **Save** button.
After saving, the context appears in the cards. Click **View** (or click the card) to see the full content. Click **Edit** to make changes.
Describe what you want to learn from the interviews. Include:
* Research objectives
* Questions you want answered
* Hypotheses to test
* Product or service context
* Any specific focus areas
Provide the structure of your interviews. Include:
* Opening questions
* Main discussion topics
* Follow-up questions
* Closing questions
```
Product Description:
A mobile banking app with instant transfer feature.
Research Goal:
Understand customer satisfaction with the new feature, focusing on challenges and emotions.
Additional:
Interested in frustrations and positive feelings while using the feature.
```
```
Opening:
- Thank you for participating
- Explain purpose and duration
- Confirm recording permission
Main Questions:
1. Tell me about your experience with [feature]
2. What challenges do you face?
3. How do you currently solve this?
4. What would make this easier?
Closing:
- Any additional thoughts?
- Thank you for your time
```
## Documents (Coming soon)
The Documents card is currently in development. This feature will allow you to upload supporting documents that provide additional context for analysis.
## How context is used
Research goals and interview guide are used as the default context for AI interview links in this project.
Research goals and interview guide are automatically included when generating reports.
Context is applied to all records in batch operations.
Context helps synthesize insights across multiple interviews.
Better context produces more relevant insights.
## Best practices
Write goals as outcomes, for example "Understand why users abandon onboarding" rather than "Learn about onboarding."
Keep interview guides structured but flexible. Include main topics and example questions, but allow for natural conversation flow.
Update context as your research evolves. If your goals change, update the Research Goals to reflect the new focus.
Use [Interview Guide Builder](/interviews/guide-builder) to create structured guides for your research. You can save AI-generated guides directly to the project.
For new projects, use the Research Assistant to generate a research context brief. The assistant guides you through required sections and creates a document you can save directly to the project.
Select **Fast** mode (2 tokens) for quick responses and standard brief generation. Use **Thinking** mode (5 tokens) when you need deeper analysis for complex research scenarios. You can switch modes at any time during the conversation using the mode selector in the input area.
If you have existing research materials, documentation, or notes, you can provide them in your first message. The AI will analyze the materials and extract information, asking only follow-up questions about missing or unclear details. This can save time compared to answering questions one by one.
If you don't know the answer to a question, click the **Skip** button on the question card. The AI will ask alternative questions to gather the information it needs. Skipping is useful when you haven't decided on a particular aspect yet or the question doesn't apply to your research.
You can ask the assistant to modify an existing research context brief. For example:
* "Add more details about hypotheses"
* "Update the participant segments"
* "Change the analysis approach to focus on JTBD"
* "Make the brief shorter"
The AI will regenerate the brief based on your modifications.
The assistant uses validation to ensure your research context brief is complete. If important information is missing, the AI will ask follow-up questions with different priorities:
* **Must** questions: Critical information needed before generation
* **Should** questions: Helpful but not required
The AI prioritizes "must" questions and won't ask for confirmation to generate until all critical information is collected.
If there's an error during generation (e.g., LLM service unavailable), the system uses a fallback template to create the brief from collected data. Fallback documents are marked with "(fallback)" in the title and include a warning message. While functional, fallback documents may lack some contextual refinement. You can regenerate the brief when the service is available.
## Force specific tool calls
You can force the Research Assistant to use a specific tool for your request. This ensures your message is processed with the right context and capabilities.
You can select an action in two ways:
**Option 1: Use the action button**
* Click the **`+`** button in the input toolbar
* Select an action from the dropdown menu:
* **Research Assistant**: Force the assistant to create research context briefs
* **Interview Guide**: Force the assistant to generate interview guides
* **Ask Project**: Force the assistant to answer using project documents (with citations)
* **Documentation**: Force the assistant to answer questions about Aseed
**Option 2: Use slash commands**
* Type **`/`** in the message input
* A menu appears with the same action options
* Select an action to apply it to your message
The selected action appears as a badge next to the input. You can remove it by clicking the badge.
Type your message and send it. The AI will process your request using the selected tool, ensuring the response is tailored to that specific function.
Forcing a tool is especially useful when you want to:
* Ensure your research context request uses the Research Assistant tool
* Make sure your guide request uses the Interview Guide tool
* Get answers grounded in project documents using Ask Project
* Get documentation-specific answers using the Documentation tool
When an action is selected, messages are marked with a badge indicating which tool is being used (e.g., "Context Builder", "Guide Builder", "Ask Project", "Documentation").
## Ask questions about Aseed
The Research Assistant can answer questions about Aseed features, workflows, and best practices. Click **"Ask About Service"** in the empty state or ask questions like:
* "How do I create a project?"
* "What report types does Aseed support?"
* "How does batch generation work?"
* "How do I export reports?"
You can also force documentation mode by selecting the **Documentation** action before asking your question. This ensures the AI uses the documentation tool to provide accurate, up-to-date answers.
Messages with documentation answers are marked with a **"Documentation"** tag to indicate they're based on Aseed documentation.
## Related topics
* [Create a project](/projects/create)
* [AI Interviewer](/projects/ai-interviewer)
* [Create interview guide](/interviews/guide-builder)
* [Upload records](/projects/upload-records)
* [Batch generation](/projects/batch-generation)
* [Generate single reports](/reports/single)
* [Create project reports](/reports/project)
* [Token usage](/account/usage)