Case Study · Law Firm · The result, in one breath
Cowboy Legal's 150-page PDFs, handled end-to-end. A paralegal afternoon, replaced by six Make.com scenarios that fire on intake and finish before the coffee's done.
A 150+ page PDF lands in the firm's intake every week. Before automation, a paralegal would split it by hand, re-key key fields into a form, collate the output, chase a signature, and email the client a summary. Accurate work. Expensive work. Not the work anyone went to law school for.
Here's what happens now, in the time it takes to make coffee. Each section below is one live Make.com scenario — six in all, chained. Watch the canvas behind this text build piece-by-piece as you read.
A new PDF hits a Gmail-monitored intake. Make.com catches the webhook, fetches the attachment via HTTP, and starts building the folder scaffold in Google Drive — month folder, temp Wyodocs folder, temp invoices folder, and more — all deterministic, all named the same way every time.
A Router then hands the file off to PDF.co to split it into its component documents. An Iterator walks through the results and uploads each split chunk to the temp invoices folder. Every ID lands in the Data store for the next scenario to find.
Scenario 2 fires once per split chunk. It downloads the temp invoice from Drive, sends it to PDF.co for text extraction, then pipes the extracted text into two OpenAI calls: one pulls structured invoice data (parties, line items, amounts), the other pulls start/end dates. Both answers land back on the same Data store record.
This is where the PDF stops being a PDF and starts being structured data — ready for the form fill in the next scenario.
Now we have clean structured data and a PDF template with empty fields. Scenario 3 looks up the record in the Data store, calls PDF.co's "Fill a PDF Form" endpoint to populate the template, uploads the filled document back to Drive, and updates the record with the new file ID. Seven steps, all automated. Zero copy-paste.
The filled PDF form has interactive fields — not what we want to ship to a client. Scenario 4 downloads the un-flattened version, converts it to images via PDF.co, reassembles it as a flat PDF, and uploads the clean deliverable to Drive. The record updates with the flattened file ID.
Time to merge. Scenario 5 pulls every invoice and every Wyodoc for this intake, feeds them through PDF.co's combine step (with Sleep/Job Check waits because PDF.co jobs are async), uploads the combined final package to the client's delivery folder, then cleans up the temp folders — both temp Wyodocs and temp invoices — so only the deliverable remains.
This is the longest scenario by node count. Also the most satisfying — the workspace is tidy when it's done.
The finale. Scenario 6 pulls every record for this intake, writes each row into a Google Sheet summary, runs math expressions for totals and counts, updates the summary row, downloads both the summary and the final combined PDF, and emails the client — attachments included.
Last step: delete all the Data store records for this intake. The pipeline is clean for the next one.
Tell us the document workflow. We'll tell you, honestly, whether automation can fix it — and what it would cost to find out.
Schedule a free workflow audit →