Case Management OCR: Getting Scans Into a Matter File

Case management OCR means giving every scanned paper a text layer before it is filed to a matter, so the system's search can find it later. Convert the scan to a searchable PDF, name it to a fixed pattern, attach it to the matter and index it. A document filed without text can only be found by its name.

Case management software stores whatever you hand it. Whether it can also look inside a scanned page depends on the product, and the only way to know is to test yours. The steps below start with finding that out, then cover the order of work, the names, the incoming paper that most needs converting, and the dates.

What does OCR add to a case management system?

OCR adds the words. A case management system files a document under a matter with a name, a date and perhaps a category. With a text layer behind the scan, the document also answers a search for anything written on its pages, across the whole matter or the whole firm.

Think about what you actually search for a year into a case. It is rarely "the March letter". It is a doctor's name, a claim number, the phrase "without prejudice", or the amount of a settlement offer. Those live on the page, not in the file name.

A text layer also makes a document reusable. You can copy a paragraph from an order into a letter, paste a set of requests into a response template, or pull a schedule into a spreadsheet. Without text, each of those is retyping.

What OCR does not add is structure. It will not tell the system that a document is an answer, who filed it, or which deadline it triggers. Those are still fields a person fills in. And recognized text is a reading of a picture, so it can be wrong in exactly the places you care about. The later section on dates comes back to that.

Does your case management software search inside scanned PDFs?

Find out with a five-minute test instead of the sales brochure. File two versions of the same page to a test matter, one image-only and one searchable, wait for indexing, and search for a word that appears on the page but not in either file name.

The test separates three situations that look alike from the outside:

1. The system searches names and fields only. Neither file is found. Text layers will not help search inside the system, though they still help once a document is open.
2. The system searches text that is already in the file. Only the searchable version is found. It means every scan you file without OCR is invisible to search.
3. The system runs its own recognition on what you file. Both are found, perhaps after a delay. You may not need to convert first, but you should check how good its text is.

Running the test

1. Pick a typed page with an unusual word on it, such as a surname or a product code.
2. Scan it, or take an existing image-only scan. Confirm you cannot select the text in a PDF reader.
3. Make a searchable copy of the same page with any OCR tool.
4. Name the files so that neither contains the word: test-A.pdf and test-B.pdf.
5. File both to a matter set up for testing.
6. Search for the word straight away, after an hour, and the next morning. Indexing may not be instant.
7. Write down which files came back and when.

Repeat with a page of poor quality, such as a fax. If the system does its own recognition, this shows you how it copes. Also try searching from wherever your staff really search: the matter view, the global search box, and the mobile app can behave differently.

Keep the result somewhere people will see it. A one-line note in the office procedures, "Our system only searches text already in the PDF, so OCR every scan before filing", prevents a year of unfindable documents.

File the scan under the matter name your docket already uses. While a motion is sub judice, you will search for the judge, the hearing date, or the pleading title. An attorney and a paralegal both need those words inside the PDF. If the matter search misses a document you can see on screen, the text layer is the piece that is missing.

Should you OCR a document before or after filing it to the matter?

Before, in most offices. Converting first means the version attached to the matter is the searchable one from the start, with its final name. Converting afterwards leaves two copies to reconcile, and the image-only one is usually the one people keep opening.

The order that works for incoming paper:

1. Scan. One document per file. Settings matter more than software here, and the section on scanner settings that make OCR work covers them.
2. Convert. Run OCR and export a searchable PDF, which keeps the page image exactly as scanned and puts text behind it.
3. Check. Search the PDF for a word you can see. Read the names, dates and amounts on the first page against the image.
4. Name. Apply the pattern in the next section.
5. File. Attach the searchable PDF to the matter and fill in the system's fields.
6. Index and calendar. Add the row to the pleadings index and send any dates for review.

There are fair reasons to file first and convert later.

• Proof of receipt. Some firms want the scan filed the minute it arrives, so the system's timestamp shows when the paper came in. File the raw scan, then add the searchable version as a new version of the same record, not as a second document.
• A backlog. Years of image-only documents are already in the system. Converting them is a project with its own plan, starting with open matters.
• Urgent paper. A notice with a hearing date goes to the responsible lawyer first. Conversion can wait an hour. Calendaring cannot.

If you do convert after filing, decide what happens to the first copy. Two files with the same name and different contents will confuse everyone. Replace, version or clearly rename, and write down which.

One point about where the conversion happens. A scan sent to an outside OCR service leaves the case management system on the way. If your matter documents are confidential, which they are, check the service's terms first. Schema Law states its commitments under security on the home page: documents are never used to train an AI model, stay inside your own account, are encrypted in transit and at rest, and can be deleted at any time.

Naming converted pleadings and motions for a matter file

Name a filed paper by its date, the party that filed it and its title as captioned, in that order, and keep a pleadings index that lists every filing in sequence. The name lets a person recognize the document in a list. The index shows what is missing.

A pattern for court papers:

YYYY-MM-DD_Party_Title-as-captioned.pdf

2026-02-03_Pltf_Complaint.pdf
2026-03-10_Def-Brightwater_Answer-and-Affirmative-Defenses.pdf
2026-04-22_Court_Scheduling-Order.pdf
2026-06-01_Def-Brightwater_Motion-to-Compel.pdf
2026-06-15_Pltf_Opposition-to-Motion-to-Compel.pdf

Some choices inside that pattern are worth making on purpose.

• Which date? Use the filing date for filed papers and the service date for papers served but not filed. Whichever you choose, use it every time.
• Docket numbers. If your court numbers docket entries, put the entry number after the date. It is the one label the court, the other side and you all share.
• Converted copies. When you also export an editable Word version to draft from, give it the same base name plus a suffix such as _working. Nobody should mistake it for the filed paper.
• Exhibits. File long exhibits as separate documents named after their parent, so a search hit lands on the exhibit and not on page 214 of a motion.

The pleadings index

The index is a table with one row per filing. A system may generate a document list, but a list sorted by upload date is not an index. Keep one that follows the court file.

The last column is the one this post adds. It records whether the filed copy passed a search test, and it turns "we OCR everything" from an intention into something you can audit. A gap in the numbers tells you a filing never reached the matter.

If several people file to the same matters, a shared structure matters more than any one name. Keep that structure in the case management system, where everyone already works. An OCR account will not supply it: in Schema Law each account belongs to one person, and no other account can see its documents. The post on OCR for a mid-size law firm covers how a larger practice divides the work.

Served papers and court notices on the way into a matter

Convert everything that arrives as paper or as an image-only attachment and that someone will need to search, quote or act on. In practice that is served pleadings and discovery, court orders and notices, and correspondence from the other side. Sort the day's intake by what each item triggers.

Two of those rows have their own guides. Requests served as a scan need careful handling of numbering before you draft, which the post on converting interrogatories from PDF to Word walks through. A judgment has figures that must be exact, covered in the post on OCR for a judgment PDF.

Email attachments deserve the same suspicion as paper. A PDF attached to an email from another firm may be a scan of a printout. Select a word in it before you file it. If nothing highlights, it goes through conversion like anything from the copier.

For the conversion itself, Schema Law takes a PDF, PNG or JPEG of up to 20 MB in the browser and shows the parsed pages before you export. The free plan covers one document a day of up to three pages, which suits a short notice. The pricing page lists what the paid plans allow.

Pulling dates out of converted orders for the case calendar

Use the converted text to find every date in an order, then have a person read each one against the page image and work out what it means before anything goes on the calendar. OCR can locate dates quickly. It cannot be trusted to transcribe or interpret them, and Schema Law does not identify deadlines for you.

Finding the dates

Once an order is searchable, search it for the things that signal a deadline:

• Month names, and the year in digits.
• "within", "no later than", "on or before", "days".
• "hearing", "conference", "trial", "shall file", "shall serve".

If the order contains a schedule laid out as a table, an Excel or CSV export gives you its rows and columns to paste into a working list. Treat that list as a draft.

Why a person must check

A date is a short string of digits with no spare characters. "June 18" read as "June 13" is still a plausible date, so nothing looks wrong. Schema Law's output does not mark a passage as uncertain, and no tool's warnings are a substitute for reading the page. Stamps, faint copies and handwriting over a typed date make misreads more likely, and judges do write changes by hand.

Even a correctly read date needs counting. Under Rule 6 of the Federal Rules of Civil Procedure, a period stated in days is computed by a set method: "exclude the day of the event that triggers the period", count every day, and if the last day is a Saturday, Sunday or legal holiday, run to the next day that is not. The same rule adds time after certain kinds of service. That is the federal civil rule only. State courts, local rules and individual judges' orders have their own methods, so use the one that governs your matter.

A calendar entry sheet keeps the check honest:

Two people should appear on that sheet for anything that could forfeit a right: one who reads the page, one who enters the date. The converted text gets you to the right paragraph in seconds. From there, the checking belongs to people. How the documents reach text in the first place is shown in the how it works section.

For one system in particular, see OCR for Clio, which covers what Clio documents about search and how to test it.