The report section helps you create a comprehensive report based on your input datasets. You can access this feature in the “Report” section of the “Generate” view.

You have four options for creating the report sections, each represented by a card. Keep in mind that you can edit all prompts, even when using templates. If you are a first-time user and are unsure about the type of report you want, we recommend starting with either the “Interest-based report template” or the “Landscape-based report template”. Select a card to get started:

New blank report template: This option gives you complete flexibility over the prompts, sections, and style instructions, allowing you to define the report sections from scratch.

Import existing report template: After creating a report using any of the report cards, you can save it as a template for reuse with other datasets. This option lets you use an already-created template. To learn about report templates, refer to the article Creating a template for auto-generating a report.

Interest-based template: AI will automatically prefill the report sections based on your specified area of interest.

Landscape-based report template: AI will prefill the report sections based on the content landscape available in the “Explore” view.

To exemplify the different components of the report, we selected “Interest based report template” and specified the area of interest as “Developments in the electric vehicle industry”.

Each row with a number next to it represents a different section of the report, while each column indicates a specific component of the report. You can rearrange the order by hovering over the number to the left of the first column and dragging the row to your desired position in the report. Now, let’s go through the columns.

Title: Enter the section title for the report. Use “/” as a hierarchy marker; for example, in the screenshot below, the first row has the section title “Market Trends in Electric Vehicles” with the sub-section title “Consumer Adoption Rates”. The second row is under the same section with a different sub-section title: “Market Growth Projections”. You can add more “/” characters to the title to create additional subsections within a row. Title is a mandatory field.

Knowledge Bases: Include the Knowledge Bases you want the AI to use when generating a response. You must select at least one Knowledge Base from the dropdown list, as the selected Knowledge Base(s) will serve as input for the report. Knowledge Bases is a mandatory field.

Table inputs: If you have any purely quantitative information in an Excel table, you can add it by clicking the bar chart icon at the bottom of the page and selecting the relevant table from the dropdown in “Table Inputs.” Ensure that the column headings are clear, and consider providing contextual information above the table, in cell A1, to improve clarity. Note that Table Inputs is an optional field.

Instruction: Describe what the AI should do with the provided Knowledge Bases and the specific question you want answers for. Focus on the core task, and avoid instruction on response style, tone, or length, as those will be specified in the 'Style Preferences' section.

For example, instead of asking, “What factors influence consumer adoption of electric vehicles?” you might say, “Describe the factors influencing consumer adoption of electric vehicles, including any relevant numeric information or figure”. Instruction is a mandatory field.

Scroll to the right side to see the remaining columns.

Example output: If you want your output to mimic a specific style, you can add it here. Example Output is an optional field.

Style preferences: If you have specific formatting, tone, or style preferences for a section, include them here as clear instructions, similar to the ones in the example. If your preference applies to all sections, specify it in the pipeline sidebar, as explained below. Style Preferences is an optional field.

Click the arrow on the right side of the page to open the pipeline sidebar, which will reveal additional settings.

Default style preferences: This section allows you to set style preferences that apply to all sections. Instead of repeating the same style preference for each row, you can enter it here, and it will automatically be applied to every section.

Placeholders: Placeholders are useful for reusing the same instruction across different sections while changing specific elements. For instance, if you are creating an overview of developments related to various companies -where each company has its own section or subsection but the prompt remains the same- you can use placeholders to represent these companies in your instruction. Define a placeholder under a specific name, such as "company1", and include it in your instruction using the format $company1. This allows you to swap the placeholder with different company names while keeping the rest of the prompt consistent. When you successfully use a placeholder in your prompt, it will be highlighted as long as the “Highlight Placeholders” configuration is enabled. Placeholders are optional.

Section settings that apply to all sections will open when you click “Open”.

Batch update allows you to apply specific adjustments to all sections of a report simultaneously, eliminating the need to change them one by one.

Knowledge Bases: You can update the Knowledge Bases across all sections at once, ensuring that any changes are applied universally in a single step.

Table inputs: Similarly, you can update the table inputs across all sections at once, guaranteeing that any modifications are applied universally in one action.

Include references: You can choose whether to include references. When enabled, each section output will feature references in square brackets within the text and listed underneath.

Knowledge base order sensitive prioritization: When this option is enabled, the sequence of Knowledge Bases matters for the AI when drafting a response. The more important a Knowledge Base is, the earlier it should be added.

LLM Engine: In most cases, you can keep the default engine. For more complex instructions, you may switch to a more advanced LLM engine, which is expected to follow instructions more diligently.

LLM Mode: This refers to the level of creativity applied by the LLM, with three main settings: "Creative," "Balanced," and "Precise." These options control how the LLM generates responses, ranging from most to least creative in that respective order. This is similar to adjusting the temperature setting in LLM models.

Output Mode: There are two available output modes: regular and comparison.

Icon 1 edits the overall title of the report.

Icon 2 adds a new section to the report, which translates into a new row in the interface.

Icon 3 duplicates the instruction and settings for the selected section in a new row.

Icon 4 enables you to configure the following settings:

Output Mode: There are two available output modes: regular and comparison.

Icon 5 takes a run of the selected section with your configurations. You can adjust the settings and prompt, evaluate the results, and iterate as needed to achieve the desired outcome.

Icon 6 deletes the section.

Icon 7 deletes the report.

Icon 8 allows you to add and remove tables.

Icon 9 converts your report into a template that you can use later in future Insight Booster projects. To learn about report templates, refer to the article Creating a template for auto-generating a report.

Icon 10 generates and auto-downloads the full report as a Word document.