Contract OCR at Signing: An Intake Routine for New Deals

Run contract OCR on the day the signed contract comes back, before the file is closed. Confirm the scan is whole and fully signed, convert it, read the parties, dates and amounts against the page images, then file the untouched scan and its searchable text together under one name with a short record of what you checked.

Signing day is the one point at which every question about the scan still has an easy answer, which is the whole case for doing the work then.

When in a contract's life should OCR be run?

Run OCR at signing, as part of closing the file, and not years later when a dispute, an audit or a sale forces someone to read the contract. On signing day the signers can still supply a missing page, the paper is still on a desk, and the negotiator can still confirm what a blurred figure says.

A contract passes through a few moments at which someone could convert it. Compare what each one offers:

Three things are true only at the start.

A defect can be cured by asking. A cut-off page, a missing exhibit or an unsigned signature block is an email on signing day. Whether it can be cured at all three years on depends on people and paper you no longer control.

A known-good text exists to check against. The final draft that went out for signature is still in your sent mail. Compare the scan's page count and section headings with it while you know which draft was final.

Search works from the first day. A renewal date, a consent requirement or a cap is findable by anyone in the team from the moment the file is closed, without waiting for a project.

If your problem is the reverse, a cabinet of old paper agreements that were never converted, the order to attack it in is covered in the guide to which legacy contracts are worth scanning and converting first. The routine here is for the agreement signed this week.

What arrives on signing day: ink-signed scans, phone photos and e-signed PDFs

A newly signed contract reaches you in one of three forms: a scan of ink-signed paper, a photo taken on a phone, or a PDF signed electronically. The first two are pictures of pages and need OCR. An e-signed PDF may already hold real text, so test it before you convert anything.

Ink-signed paper, scanned

Someone printed the agreement, signed it and fed it through an office scanner or a multifunction copier. The result is an image-only PDF. Every page is a picture, so nothing can be selected or searched until recognition adds text. Watch for two-sided originals scanned one side only, and for signature pages scanned separately and sent as a second attachment.

A phone photo or a scanning app

A signer away from the office photographs each page, or only the signature page, and sends JPEG files or a PDF made by an app. Expect shadows, curved pages and pages in the order they were photographed. If only the signature page was photographed, you hold a signature page and an unsigned draft, and you need to decide today whether that is the record your team keeps or whether to ask for the whole document.

Treat the signed set as the writing you will describe if a statute of frauds question comes up. Record who signed, the date on each block, and which pages are actually in the file. Leave a missing page as a blank line in that list. The lawyer decides what the gap means.

An electronically signed PDF

A signing service returns a PDF that was built from the word-processor file. Try to select a single word on a middle page. If the word highlights, the text is already there and running OCR adds nothing. Keep the file exactly as delivered, with any completion certificate, because the post on what is inside an electronically signed PDF besides the signature explains what a print-and-rescan would discard.

Mixed sets

One party signs electronically and the other prints, signs in ink and scans. You then hold a text PDF from one side and an image PDF from the other. Sort what you received before you start:

A signing-day contract OCR routine in seven steps

The routine is seven steps: save the file as received, confirm it is complete and fully signed, convert it, proofread the terms that carry risk, record the key details, file scan and text together, and report anything open. Do them in that order, because each step relies on the one before it.

The fictional example is a supply agreement between Brightwater Foods, Inc. and Kessler Packaging LLC, signed on 6 October 2026 and returned by Kessler as a 24-page scanner PDF.

1. Save the file exactly as it arrived. Put the attachment in the matter folder before opening it in any editor. Record who sent it, when, and its file name. This copy is never renamed, edited or resaved.
2. Confirm the scan is complete. Count the pages against the final draft and check that every exhibit named in the table of contents is present. The tests for page numbers, footers and section sequence are set out in the post on missing pages in a scanned contract.
3. Confirm it is fully signed. Look at every signature block on the page images, not in converted text. For Brightwater and Kessler that means two blocks on page 22 and an initialed price exhibit on page 24. Note any block that is blank, dated differently or signed by someone other than the named signatory.
4. Convert a working copy. Upload a copy of the scan and export a searchable PDF, plus a Word or plain text file if someone will quote from the agreement. In Schema Law you upload a PDF, PNG or JPEG of up to 20 MB from the browser and see the result page by page on screen, so you can check it against the original, and the export formats on the home page list what each file type contains.
5. Proofread the terms that carry risk. Read the party names, the effective date, the term, every amount and every notice period against the image. Nothing in recognized text marks a doubtful character, so the comparison is yours to make.
6. Record the key details in an intake sheet. One short form per agreement, kept with the file.
7. File the scan and its text together, then report. Use one base name for both files, put them where the team's contracts live, and send the deal lead a note listing anything still open.

A filled intake sheet for the example:

When an agreement is one of many signed at the same closing, the single-document routine still applies to each, and the indexed set of the whole transaction is the subject of the post on building an electronic closing binder.

Which contract intake checks must happen now and which can wait?

Do today the checks that only get harder with time: completeness, signatures, a text layer, and the proofreading of parties, dates and money. Leave for later anything that needs only the converted text and a quiet hour, such as a full clause summary or a comparison against your template.

The test for each task is one question: if this turns out to be wrong, who do you need in order to fix it, and are they still available next year?

Two cautions on the "can wait" list.

Deferring a task is a decision to record. Write "full proofread not done" on the intake sheet. A Word file with no such note looks checked to the next reader.

Dates do not wait long. Turning a notice period into a calendar entry has its own traps, and the method is in the post on how to turn a contract notice period into a calendar date. Do it in the first week while the deal is fresh.

A second fictional case shows the triage. Tamsin Okafor, the only paralegal in a four-person legal team, receives a distribution agreement at 4:40 on a Friday as eleven phone photos. She orders the images, sees that page 7 is absent, and emails the signer before leaving. On Monday she has page 7, converts the set, and proofreads the territory, the minimum purchase figures and the term. The clause summary the sales director asked for waits until Wednesday. Nothing was lost by waiting for the summary. Page 7 could have been lost by waiting.

Naming and filing a newly executed contract with its text

Give the scan and every file made from it the same base name: the date on the agreement, the counterparty, the agreement type and a status word. Store them side by side, with the scan as received kept apart and unchanged, so nobody has to guess which file is the signed record and which is working text.

A pattern for a newly signed agreement:

2026-10-06_Kessler-Packaging_Supply-Agreement_EXECUTED.pdf
2026-10-06_Kessler-Packaging_Supply-Agreement_EXECUTED_searchable.pdf
2026-10-06_Kessler-Packaging_Supply-Agreement_EXECUTED_text.docx
2026-10-06_Kessler-Packaging_Supply-Agreement_intake.xlsx

The parts and the choices behind them:

• Date first, year first. Use the date written on the agreement. If the parties signed on different days and the agreement states no effective date, pick one rule for the whole team, such as the last signature date, and write the rule down.
• Counterparty, not your own client. In a legal department every contract has the same first party. The other side's name is what people search for.
• The agreement's own title. If the document calls itself a "Supply Agreement", do not file it as a "vendor contract". Search has to match the words on the page.
• A status word in capitals. EXECUTED for fully signed, PARTIAL when a signature or initial is outstanding. Rename from PARTIAL to EXECUTED only when the complete version arrives, and keep the partial one.
• A suffix for each derived file. _searchable, _text, _tables. The suffix tells a reader the file was made from the scan and is not the scan.

Where the files go

Keep three things apart inside the contract's folder:

1. As received. The email attachment with its original file name, untouched.
2. Record copy. The named scan and its searchable twin.
3. Working files. Word, text and spreadsheet exports, the intake sheet, and later summaries.

The Word export deserves a line at the top saying it is converted text and not the signed agreement. Why that label matters, and what it should say, is covered in the post on why the OCR text of a contract is never the contract.

If the conversion was done in an online service, decide on the same day whether the uploaded copy stays there. Schema Law's pricing page states for every plan "Delete any document, at any time", and nothing is deleted for you, so a copy stays until you remove it. Your record is the copy in your own filing system either way.

Who owns contract intake in a small legal team?

One named person owns intake for each agreement, and in a small team that is whoever closes the file: a paralegal, a contracts manager or the lawyer who ran the deal. The owner need not do every step, but they sign the intake sheet and chase the open items.

The risk in a small team is easy to picture. The lawyer treats the deal as done when the signatures arrive, the business contact assumes legal has the contract, and the scan sits in two inboxes. A named owner and a written handoff close that gap.

One way to divide the work among three roles:

In a team of one, every column is you. The table still helps, because it lists the hats to put on in turn.

A handoff checklist

The deal lawyer sends this with the signed file. The intake owner returns it completed.

• Which draft was final, with its file name and date.
• Who signed for each party, and how each signed: ink, electronic, or both.
• Whether any page was changed by hand at signing.
• Which exhibits and schedules should be attached.
• Where the paper original is, or who is sending it.
• Any date the business contact has already been promised a reminder for.
• Who may see the agreement inside the company.
• Open points the lawyer already knows about.

The third item deserves care. A handwritten change made at the signing table is invisible in a list of terms taken from the final draft, and OCR in general is weaker on handwriting than on print. If the lawyer answers yes, the intake owner follows the method in the post on handwritten changes to a contract before treating the converted text as complete.

Set one expectation in writing for the team, such as intake finished within two working days of the last signature, and review the open items list once a month. A routine this small survives only if someone is asked whether it happened.