Back to previous work

Case Study

AI Contract Briefing & Change Monitoring System

A document workflow system that parsed contracts from email, generated briefing reports for attorneys, and monitored contract changes that could affect a client's position.

Legal · Contract Review · Briefing Reports · Change Monitoring

AI contract briefing and change monitoring system preview.

Challenge

The challenge was producing attorney briefings from a high volume of incoming contracts without a paralegal building each one from scratch.

System

A workflow system read contracts from email, generated a briefing report for each one, compared it against the existing version on file, and presented it to a paralegal for review and approval before it reached the attorneys.

Outcome

Briefing prep became 2.5x faster, time per document fell from 25 minutes to 10, about 6,250 paralegal hours were returned each year, and contracts never left the firm's environment.

Background

The client was a law firm that received contract documents by email, on the order of 500 a week. Attorneys discussed each contract from a briefing report, and paralegals prepared those reports.

The workflow was designed around paralegal review, keeping human approval in place before attorney-facing briefings were delivered.

Problem

Around 500 contract documents arrived each week, and a paralegal had to read each one and build a briefing report for the attorneys by hand, which took about 25 minutes per document.

Some contracts also had to be checked against an existing version held in the firm's internal tool to flag changes that could affect the relevant client's position, which meant locating the prior contract and comparing the two clause by clause.

Incoming contracts had to be briefed and, when a prior version existed, compared against the version already on file.

How it was solved

The system read incoming contracts from email and extracted their text, with OCR for scanned or image-based documents, so the contents of each contract were available to the model.

For each contract, it generated a briefing report covering the terms an attorney needed, such as the parties, key dates, obligations, and provisions material to the client. Every item in the briefing was linked back to the place in the document it came from, so a reviewer could check a flagged term against the exact clause behind it.

When an incoming contract matched an existing one held in the firm's internal tool, the system compared the two and flagged the changes, with attention to the terms that could affect the relevant client's position.

A paralegal reviewed each briefing in a front end, edited it where needed, and approved it before it reached the attorneys. The paralegal stayed accountable for what went to the attorneys, and the source links let them verify each item against the contract quickly.

The system ran on the firm's own AWS infrastructure with a self-hosted open-weight model, so the contracts never left the firm's environment. It processed documents through a job queue so a burst of incoming email did not stall briefing prep.

Impact

  • 2.5x faster briefing prep. Time to prepare a briefing fell from 25 minutes per document to 10.
  • 6,250 paralegal hours returned a year. Across 500 documents a week, the team got back about 125 hours weekly.
  • Contracts never left the firm's environment. The system ran on the firm's own infrastructure with a self-hosted open-weight model.

Briefing prep ran 2.5x faster, from 25 minutes per document to 10. Across 500 documents a week, that returned about 125 hours of paralegal time weekly, on the order of 6,250 hours a year. Split across five paralegals, each got back around 25 hours a week.

The dollar value of that time depends on the loaded paralegal rate, so the table below shows it across a range rather than fixing a single figure.

Loaded paralegal rateTime returned per year
$40/hr~$250,000
$55/hr~$344,000
$70/hr~$438,000
$85/hr~$531,000
These figures show the value of the time returned to the team. The firm realizes it as paralegal hours redirected to other work.