When you import a PDF, AI reads the contents and automatically builds an SOP. You can reuse existing manuals and work standards as Dive SOPs. AI also creates the title, body text, tips, cautions, and safety notes for each step from the document.

About creating an SOP from a PDF

This method has AI build an SOP from the imported PDF. The following content is generated:

Overall title and summary of the PDF
Division into steps (by default, table rows and bullet points within a page are also split into steps by content)
Title, body text, tips, cautions, and safety notes for each step
Questions (when [Create questions] is selected in the AI options)
Each step displays the corresponding page of the original PDF as is

By default, the body text is organized into easy-to-read sentences. If you want to keep the original wording, select [Transcribe] for [Body writing style] in the AI options. A single step never spans multiple pages.

Preparation (supported file formats)

In addition to PDF, you can import Excel, Word, and PowerPoint files as they are (there is no need to export them to PDF first). For details on importing Office files, see Creating an SOP from a file (Excel/Word/PowerPoint supported).

Note that PDFs with selectable text (PDFs that contain text data) produce more accurate results than PDFs that are simply scanned images.

How to create

When creating a new SOP

From the sidebar, click [+] → [Register existing SOPs], then click [Choose from materials] at the bottom of the screen that opens.
Upload a PDF file, or select an AI-processed PDF from [Materials] (if you upload a file, set the AI options and run processing. See "Setting AI options" below).
Under the heading "View as a Dive SOP" in [How do you want to import the file?], select an import method from [Whole file as 1 step] / [1 step per slide] / [Select slides]. (Selecting [File as is] at the top creates an SOP that shows the PDF as is, without splitting it into steps. For the differences between the options, see "How to choose an import method" in Creating an SOP from a file (Excel/Word/PowerPoint supported).)
Check [Step conversion settings], then click [Confirm with this content].
A draft is created, so check the contents, edit as needed, and publish.

When adding steps to an existing draft

You can also import a PDF and add steps from the SOP editing screen. Use this when you want to insert the contents of a PDF into an existing SOP. When adding a step, select [File] and then select a PDF. You can then choose an import method from [Whole file as 1 step] / [1 step per slide] / [Select slides]. For PDFs that have not been processed by AI, only [Whole file as 1 step] can be selected.

Setting AI options

When you upload a PDF, the AI processing options open. Hover your mouse over the description icon next to each item, or over the option buttons, to see an explanation. The settings you chose last time are already filled in.

Analysis instructions

Briefly telling the AI what kind of PDF it is will improve the accuracy of titles and body text. A description specific to this file is enough. Common premises, such as internal technical terms or industry details, can be registered once in Terminology Glossary (terms and multilingual translations), and they will be reflected in analysis without being entered each time.

Step split method

Per content (default): Splits table rows, bullet points, and multiple steps within a page, and makes each one a step. Suited to documents that have several steps written on one page
Per page: Makes each page one step. Suited to slide-style PDFs that consist mainly of diagrams or photos

Text style

Imperative (e.g., "Check the valve."): the standard for SOPs
Polite (e.g., "Please check the valve."): for training new employees, or for SOPs submitted to business partners or auditors
Suggestion (e.g., "Let's check the valve."): for education and training
Team style: If your team has defined how to write SOPs, the wording follows that definition

Advanced settings

Body writing style: [Reorganize] (default) reads the content and sorts it into work content, tips, cautions, and safety notes. [Transcribe] keeps the original wording as is and faithfully transcribes table rows and bullet points. Suited to documents that prioritize the original text, such as internal regulations, checklists, and evaluation sheets
Recognition Language: Choosing the language of the text in the PDF improves the accuracy of reorganizing and generating supplementary content
Create questions: AI creates questions for workers from the document's content. When turned on, you can select a [Question type] (On-site check / Comprehension test). When [Transcribe] is selected, this cannot be chosen, and only questions written in the original document are imported

For detailed explanations of each item, see "Details of AI processing options" in Creating an SOP from a file (Excel/Word/PowerPoint supported).

Review and edit the generated SOP

When processing is complete, a draft is created. Each generated step can be edited in the same way as a regular SOP.

Edit titles, body text, tips, cautions, and safety notes
Delete unnecessary steps and rearrange their order
Add videos, images, or AR to steps

Check the contents, edit as needed, and then publish. When editing again, you can access the target SOP from [Drafts / Approvals List].

Tips for creating SOPs successfully

Enter analysis instructions: Just adding a short note, such as "SOP for inspecting XX equipment," improves accuracy
Choose a split method that suits the document: For documents with several steps on one page, [Per content] is recommended. For documents like slides that group one task per page, [Per page] is recommended. If the steps do not split as you expected, you can change the split method and re-analyze by using [Re-run AI] in the material details
Use a PDF that contains text data: PDFs where text can be selected are more accurate than PDFs that are only scanned images

Notes and limitations

This article explains importing PDFs. Excel, Word, and PowerPoint files can also be imported from the same screen
If you want to split a PDF in [Materials] into steps, [Whole file as 1 step] / [1 step per slide] / [Select slides] cannot be selected until AI processing is complete. In that case, click [Run AI processing] in [Materials] and then try again. ([File as is] can be selected even without AI processing.)
PDFs with restrictions on printing, editing, or copying (those that do not require a password to open) can still be processed by AI as is. PDFs that require a password to open cannot be processed by AI, so please prepare a file with the password removed