Quick start: Assisted Writing
- Last Updated: November 19, 2026
- 4 minute read
- Documentation
The Assisted Writing agent assists you in creating a structured, multi-section document from the data in your project. In this quick start, you will plan, draft, review, and export a document.
This quick start guides you through the following steps:
- Start a new document
- Plan the document with an outline
- Generate draft sections
- Review the document
- Compile and export the document
For more information, see Assisted writing.
Before you begin
- Recommended: Complete Quick start: Your first project. In addition to guiding you through the prerequisites for this procedure, the quick start introduces the examples used for this topic.
- Create a project, create at dataset, and load the data that you want to use for the document. See Quick start: Your first project to guide you through those requirements.
- Confirm that the uploaded files are processed. In the dataset view, select the Uploads tab and check the status of your upload executions.
- Recommended: Enrich your data with chunking, vector embeddings, fact extraction, and classification enabled. Enabling these enrichments improves evidence discovery and can produce better results. See Enrich your resources for more information.
Example scenario
This quick start continues the scenario started in Quick start: Your first project. For this topic, the leadership team for the Manufacturing Division asks for a short report that describes the division's factories. Rather than writing the report from scratch, we use the Assisted Writing agent to create it from the factory information in the Facilities Information dataset.
1. Start a new document
When you start a document, you describe its subject and the guidance that applies to all of its content.
To start a new document:
- Open the
Manufacturing Divisionproject, and then select Assisted Writing from the sidebar menu. - Click New Document.
- Provide values for the following fields:
- Title: A name for the document. For example,
Manufacturing Facilities Overview. - Topic / Brief: A description of the document's subject, scope, and key questions. The agent uses this value to find relevant information in your data. For example,
An overview of the factories in the Manufacturing Division, including where each factory is located, what it produces, and how the factories differ in size and capabilities. - (Optional) Steering Instructions: Guidance on tone, style, and focus that applies to the whole document. For example,
Write for the division leadership team. Use a concise, factual tone, and keep each section under 300 words.
- Title: A name for the document. For example,
- Click Create Document.
The view for your new document opens.
2. Plan the document with an outline
An outline defines the sections of your document and what each section must cover. The agent creates the outline from your topic, your project data, and any outline instructions that you provide.
To create and apply an outline:
Select the Plan tab.
In the Outline Instructions field, describe how to organize the document. For example,
Organize the report into the following sections: Facilities overview, Factory locations, Production capabilities, Products by factory, and Summary.Click Generate Outline. The proposed outline appears.
Review the proposed sections. Each section includes a title, a short description, and the questions that it must answer.
Note: To change the outline, edit the Outline Instructions, and then click Regenerate Outline.
Click Apply to Draft.
The agent creates a draft section for each section in the outline. For more information, see Plan with an outline.
3. Generate draft sections
Each draft section is generated from the evidence that you select in your project data. Evidence is the text from your resources that the agent uses to write the section.
To generate a draft section:
- Select the Draft tab. The Draft view opens, and the SECTIONS sidebar lists the sections from your outline.
- In the SECTIONS list, select the first section. For example,
Facilities overview. - In the Generate panel, review the Must Answer questions for the section. Add or remove questions as needed.
- Click Find Evidence. The agent searches the
Facilities Informationdataset for information that answers each question and selects the most relevant results. - Review the selected evidence, and clear any results that are not relevant.
- Click Generate Section. The generated content appears in the Section Content area.
To generate the remaining sections at the same time, click Select & Generate, confirm the selected sections, and then start generation.
After you generate a section, you can edit its content. In the Section Content area, click Edit Markdown, make your changes, and then click Save.
For more information, see Create draft sections.
4. Review the document
The review checks your generated sections for issues such as duplicated content, contradictions, and differences in voice, and suggests changes.
To review the document:
- Select the Review tab.
- Click Run Review. The review lists any potential issues that it finds.
- For each issue, review the suggested change, and then click Apply to accept it or Ignore to dismiss it. To accept all suggestions at once, click Apply All.
For more information, see Review a document.
5. Compile and export the document
Compiling gathers the latest content from all of your generated sections into a single draft that you can export.
To compile and export the document:
- Select the Compile tab.
- Click Compile Draft. The compiled draft appears in the list of drafts.
- Optional: To preview the draft, click View.
- From the Export menu for the draft, select Export PDF.
The PDF file of your Manufacturing Facilities Overview report downloads. For more information, see Compile a draft.
Where to go next
- How to use Progress Data Platform
- Start a new document: Adjust document settings, such as the model and reasoning effort.
- Create draft sections: Refine evidence, add section steering, evaluate sections, and restore earlier versions.
- Enrichments: Enable fact extraction and classification to improve the evidence that the agent finds.