Uploading a .docx or .pdf as the source for an AI-generated SOP

Updated May 18, 2026

If you already have a Word doc or PDF describing a procedure, you do not have to retype it. MaintIQ's AI assistant can read your file and turn it into a structured SOP — but only if you tell it to. Skipping the steps below is the single most common reason the AI generates an unrelated SOP (a common report we hear: "I uploaded a water-sampling doc and got a compressor pump SOP back").

Step 1: Start a new SOP, not an AI chat

In SOPs → + New SOP, pick Start from a document. Do not pick Start with AI. The two buttons look similar but only the first attaches your file to the prompt.

Step 2: Upload the file

Drag and drop, or click to browse. Supported formats:

  • .docx (Microsoft Word, current format)
  • .doc (legacy Word — converted automatically)
  • .pdf (text-based or scanned with OCR fallback)
  • .txt, .md

Maximum file size: 20 MB per file, up to 3 files per SOP source.

Step 3: Wait for the parse to finish

After upload, you will see a Parsing… spinner. Wait until it completes (usually 5–20 seconds). The screen then shows a preview of the extracted text. If the preview is empty or garbled, the file is likely a scanned PDF with poor image quality — re-scan or retype the source.

Do not type a prompt until you see the extracted text preview. If you submit early, the AI does not yet have your document in context, and it will fall back to a generic SOP about whatever it guesses you meant.

Step 4: Write a focused prompt

Once the preview is visible, type a prompt that references the document explicitly. Good examples:

  • "Convert the attached water-sampling procedure into a MaintIQ SOP. Keep all sample IDs, hold times, and chain-of-custody steps from the original."
  • "Reformat this PDF as an SOP. Group steps under Daily, Weekly, and Monthly headings based on the frequencies in the document."

Avoid vague prompts like "make an SOP" — the AI has no way to know what part of the document matters.

Step 5: Review before publishing

The AI returns a draft. Compare it to your source side-by-side. Common edits:

  • Add missing tolerances or threshold values the AI summarised away.
  • Re-order steps to match the physical workflow.
  • Insert photos.

When the draft looks right, click Publish.

Troubleshooting

The AI generated an unrelated SOP. You almost certainly clicked "Start with AI" instead of "Start from a document", or you submitted the prompt before the parse finished. Delete the draft and restart from Step 1.

The preview shows gibberish. Your PDF is scanned and OCR could not read it. Re-scan at 300 DPI minimum, or paste the text directly.

Only the first page came through. Some PDFs use unusual layout (multi-column, embedded images covering text). Export to .docx from the source application first.