Document workflows, worked through step by step

The constraint

A proposal is never one document. It is a narrative deck from a designer, a budget sheet from finance, a scope and terms document, and two or three case studies — each exported from a different application at a different page size. Assembling them by hand is slow, and uploading confidential pricing to a third-party converter is not an option when the client has signed an NDA.

The approach

Normalise the sources before combining them, name them so the order is unambiguous, and assemble in a single pass. Because every step is a structural operation on the file, the whole workflow runs in the browser and nothing is uploaded.

Steps

1. 1.Export every source to the same page size and orientation. Fixing geometry after the merge means working page by page through a larger file.
2. 2.Rename with numeric prefixes encoding the intended order: 01_cover.pdf, 02_narrative.pdf, 03_budget.pdf.
3. 3.Verify each source individually. This is where a superseded budget gets caught, before it is buried in a 40-page document.
4. 4.Combine in one pass with Merge PDF, checking the first and last pages specifically.
5. 5.Reorder or drop pages with Organize PDF if something is out of place, rather than redoing the merge.
6. 6.Compress as a final step and compare the reported sizes against the limit of the channel you are sending through.

The constraint

Getting a manuscript from a word processor to a print-ready PDF involves repeated round trips: export, review on screen, mark up, convert the markup back into editable text, revise, export again. Each round trip is a chance to lose formatting, and print-on-demand platforms reject files for problems that are invisible in a word processor.

The approach

Keep the source document as the master and treat every PDF as a derived review artefact. Convert only when a reviewer needs to comment, and always keep the original .docx.

Steps

1. 1.Write and edit in the source word processor, where you have full control over styles and structure.
2. 2.Export to PDF for review. Do not treat the PDF as the master — it cannot be edited the way the source can.
3. 3.If a reviewer returns comments, use PDF to Word to recover the text into an editable document, then reconcile it into your master rather than editing the derived copy.
4. 4.Check the print file against the platform's requirements before submitting: page size, bleed, embedded fonts, image resolution, and colour space.
5. 5.Use Compress PDF only where the platform has a file-size ceiling, and check the result visually afterwards.

The constraint

Staff need to combine worksheets into packets, split long guides into chapters, compress files for email, and get forms signed — on a mix of managed laptops and tablets, without installing anything, and without sending student data to a third-party service.

The approach

Standardise on tools that need no installation and no account, so adoption does not depend on IT provisioning. All the operations involved are structural and run locally in the browser.

Steps

1. 1.Combine multi-page worksheets or report packs with Merge PDF.
2. 2.Split long guides into chapter-sized files with Split PDF so each one can be shared separately.
3. 3.Reduce large scanned packs with Compress PDF before emailing them.
4. 4.Place signatures on permission forms with Sign PDF. The signature is drawn onto the first page as an image, so it becomes part of the page rather than a removable annotation — but it is an image of a signature, not a cryptographic digital signature, and it only lands on page one.
5. 5.Publish the FAQ page internally as the staff reference, so the workflow does not depend on one person's knowledge.

The constraint

Redaction is the operation where a wrong assumption is unrecoverable. Drawing a black rectangle over text leaves the underlying characters fully intact in the file, where they can be recovered by copying, searching, or text extraction. Public disclosure incidents of exactly this kind are a matter of record.

The approach

Decide first whether the document must remain searchable. If it must, the content has to be removed at the stream level by a tool that does genuine removal. If certainty matters more than searchability, rasterise the pages, cover the sensitive regions in the images, and rebuild the PDF from those images — the original data never enters the new file.

Steps

1. 1.Work from a copy and keep the original untouched and access-controlled.
2. 2.Choose the method deliberately: content removal if the document must stay searchable, rasterisation if certainty is the priority.
3. 3.Handle every location, not just the visible page: metadata, bookmarks, annotations, attachments, and layers.
4. 4.Verify by selecting all, copying, and pasting into a plain text editor, then searching for the removed terms.
5. 5.Run a text extraction over the file and search that too.
6. 6.Read the document properties for author, title, and subject, and check the bookmarks pane for headings that name the redacted section.
7. 7.Have a second person verify independently. The person who performed the redaction will unconsciously see only what they intended to remove.

The constraint

Field work means capturing paper documents, applying a stamp or signature, organising multiple signed items, and delivering the result to the client — often without reliable connectivity and always without uploading client documents to a cloud service.

The approach

Photograph each document with your phone's camera, upload the images, place your signature, combine the items into one deliverable, and hand it over. Every operation runs locally, so a dropped connection does not interrupt the workflow.

Steps

1. 1.Photograph each paper document with your phone's camera, then compile the images into a PDF with Scan to PDF.
2. 2.Place your signature with Sign PDF. If you need a stamp or seal, apply it to the paper before you photograph the page — Add Watermark only draws text and has no image mode.
3. 3.Combine the signed items into a single deliverable with Merge PDF.
4. 4.Check the page order and the last page before sending, since both are easy to get wrong on a small screen.
5. 5.Rotate or crop any page that the camera captured at an angle, so the deliverable does not look like a photograph of a document.

The constraint

A literature review is a volume problem. Hundreds of documents, a mix of formats and vintages, and a genuine constraint on where unpublished or embargoed material can be sent. Older print journals arrive as scans with no text layer, so they cannot even be searched until they are processed.

The approach

Separate the local work from the model-assisted work. Do all file handling first, on the device. Then decide, per document, whether any text should go to a summarisation or translation feature.

Steps

1. 1.Triage by format: try selecting a sentence with the cursor. If characters highlight cleanly the file has a text layer; if nothing selects, it is an image.
2. 2.For scanned papers, note that this site has no OCR engine: Extract PDF Text recovers an existing text layer and tells you which pages have none. To make a scan searchable, run a dedicated OCR tool first, then check the output quality before relying on it — particularly for numbers and proper names you intend to quote.
3. 3.Build themed reading bundles with Merge PDF so each working set is one navigable document.
4. 4.Extract individual papers from large proceedings volumes with Split PDF.
5. 5.Convert only the papers you need to quote or restructure, and expect to tidy two-column layouts by hand.
6. 6.Decide separately whether any document should be sent to an AI feature, and keep a note of which ones were.