Nine hundred pages, filed the same afternoon

The organisation

Marchetti & Roe is a twelve-person immigration practice with offices in Cleveland and Akron. Four attorneys, five paralegals, two administrators and a part-time interpreter carry around forty active matters at any time. The bulk of the work is asylum and cancellation of removal, with a steady thread of family petitions underneath it.

The practice is not unusual in size. What is unusual is the shape of its paperwork: nearly everything it files is an assembled bundle rather than a written document, and the assembly is done by the paralegals rather than the attorneys.

The problem

A single asylum submission runs to roughly nine hundred pages. The I-589 and the declaration account for maybe forty of them. The rest is evidence: medical records, psychological evaluations, police reports, country-condition reports, news clippings, letters of support, identity documents, and a translator's certification attached to every item that was not originally in English.

Every one of those arrives from a different person in a different format. Medical records come as a fax the clinic printed and rescanned. Country-condition reports come as long PDFs from which four pages are wanted. Identity documents come as photographs taken on a phone, at an angle, in poor light. Roughly half of everything in the bundle is a picture of paper rather than text.

The court wants it in one file, in the order set out in the exhibit list, with an index at the front that names each exhibit and the page it starts on, and with continuous numbering across the whole bundle. Get the order wrong and the submission is returned. Get the numbering wrong and every reference in the brief points at the wrong page.

The cost of that was not the assembly itself. It was the renumbering. A medical record that arrived the day before filing — which is most of them — had to go in at its place in the sequence, which pushed every exhibit after it. In the practice's old method that meant retyping the index, restamping the pages and re-checking every cross-reference in the brief. Two paralegals lost the better part of three days to each submission, and the last of those days was almost always the day before it was due.

What they tried first

The first attempt at a fix was procedural rather than technical: a shared folder with a strict naming convention, so that sorting the folder alphabetically produced the exhibit order. It worked until an exhibit needed to go between 07 and 08, at which point the convention grew decimals, and then decimals of decimals.

The second attempt was a general-purpose PDF editor. It could merge and it could stamp, but it had no concept of an exhibit — only of pages. Inserting a document meant inserting pages, and the divider, the index entry and the number sequence were three separate manual jobs that had to be kept in agreement by hand. They frequently were not. Twice in the year before the change, a submission went to the court with an index that disagreed with the bundle behind it.

The third attempt was to freeze the bundle a week early and refuse late evidence. That lasted one matter. Late evidence in an asylum case is not an administrative inconvenience; it is often the strongest thing in the file.

What changed

The exhibit order is now a preset rather than a convention. The preset names the sections in the order the immigration court wants them, and each document is assigned to a section when it is brought in rather than renamed to sort correctly.

Dividers are generated from the exhibit list instead of typed. The index is then generated from the dividers, so the index cannot disagree with the bundle — it is derived from it. Each index entry links to the page it names, which is the part the clerks noticed first.

Translations are attached to the document they translate rather than filed beside it, so moving an exhibit moves the certification with it. That single change removed the most common defect in the practice's submissions, which was a translation that had drifted three exhibits away from its source.

Bates numbering runs last, across the assembled bundle, at the position the court asks for. Because the numbering is applied to the finished file rather than to each part as it arrives, a late exhibit is inserted where it belongs and the entire sequence renumbers itself in one pass. Photographs of documents go through OCR on the way in, so the finished bundle is searchable even where the source was not.

How it was put in

The practice did not migrate its back catalogue and did not try to change everything at once. One paralegal built the asylum preset against a matter that had already been filed, and compared the output page by page with what had actually gone to the court. That took two afternoons and found three ordering assumptions that were wrong.

The preset was then used on live matters by that paralegal alone for about a month. The other four watched the output rather than the process. When the second paralegal adopted it, the preset was already correct, so the training was twenty minutes rather than a day.

Cancellation of removal got its own preset three months later, built by the paralegal who does most of that work, from the asylum one. Nobody from the practice's IT contractor was involved at any point, which the office manager mentions more often than anything else.

The result

Assembly is an afternoon. The index is generated, so it cannot disagree with the bundle, and the practice stopped its pre-filing index check because there was nothing left for it to catch.

In the eighteen months since the asylum preset was finished, the practice has not had a submission returned for a numbering or ordering defect. That had happened twice in the preceding year, and each return cost roughly a fortnight of calendar time.

The less visible change is in when the work happens. Bundles are now assembled two or three days before filing rather than overnight before it, because a late exhibit no longer costs anything to insert. The paralegals describe that as the actual improvement, and the three-days-to-four-hours figure as the thing that made it possible.