An IP filing firm where prepping the documents took longer than the filing itself
Built a custom layer that reads the pile of documents a client sends in, renames and categorizes each one, and assembles a structured, filing-ready package with its own index, then hands clean data to the specialist filing software through its API so the same numbers are never keyed twice.
A client sends in a folder of disclosures: invention write-ups, prior-art references, drawings, assignment paperwork, all of it in whatever format the client had on hand. Before anyone can do the skilled work, a preparer has to open every file, figure out what it is, rename it to the firm's convention, group it, and build an index so the package can actually be worked. That sorting and assembly step routinely took longer than entering the file into the filing software itself, and it added nothing a client would ever see.
The firm files in two jurisdictions, and the forms and rules differ between them, so the same details often had to be entered twice into different systems. During filing-deadline season the team was buried in incoming documents and chasing missing pieces by hand, and an off-the-shelf product never fit, because the tools on the market lean to one country or the other and don't handle cross-border filings cleanly.
We built on top of the platform the firm was already standardizing on, rather than replacing it, so it slotted into the document store and the automation they already relied on. The custom layer takes the raw folder a client sends and does the work a preparer used to do by hand: it classifies each document by type, renames it to the firm's convention, groups the pieces, and produces a structured package with a generated index, every entry editable before anything is finalized.
From there it feeds the cleaned, categorized data into the filing software through its API, so a detail entered once flows to the right field instead of being retyped into a second system. The classification is AI-driven with a deterministic fallback behind it, so a noisy or unusual folder still produces a usable package and never simply stalls. We started narrow, on the roughly eighty percent of files that look the same, and left the genuinely unusual cases to a person.
The preparation step that used to eat more time than the filing now runs in the background. Documents arrive, get sorted and named and assembled into an indexed package automatically, and a preparer reviews and adjusts instead of building the package from nothing. That removed roughly forty minutes of non-billable handling from a typical file, and across the season that is the difference between a backlog and a team that keeps pace.
The double entry across the two jurisdictions is gone for the standard path: data keyed once moves to the filing software on its own. The work left for people is the judgment, checking the package and handling the odd file, not retyping and renaming. During the deadline crunch the same staff move more files without the late-night document wrangling that used to define the season.
“The bottleneck was never the filing. It was turning a client's pile of documents into something we could actually work, renaming and sorting and indexing it by hand, file after file, and the off-the-shelf tools only ever fit half of how we work.”
Have a mess like this?
Tell us where time and money are leaking. We’ll map the fastest high-ROI win, usually a Pilot that pays for itself.