Organize a clause library
Suggest a category for each extracted clause so reviewers can find related provisions.
STAYCHARTED AMT / LEGAL
Contract clause classification for legal teams
Train on clauses your team has labeled to organize contract text into your own taxonomy and direct uncertain results to review.
FROM YOUR EXAMPLES
This agreement shall be governed by the laws of the State of New York.Governing lawIllustrative category · Not a live prediction
Where this fits in your work
Prepare review queues
Group clauses by topic before a legal reviewer examines their wording and context.
Maintain a shared taxonomy
Use corrected examples to make the boundaries between similar labels explicit.
PREPARE YOUR EXAMPLES
Start with a file your team has already labeled.
Use clause text paired with a category your team agrees on: confidentiality, termination, governing law, or another label in your review process. The examples need consistent boundaries between similar categories. Decide how to treat clauses covering more than one topic before building a single-label classifier.
Use one row per example with an approved answer. Keep the input representative of what the model will receive later, and agree on how to label ambiguous cases before training.
| Clause text | Approved category |
|---|---|
| Each party shall keep the disclosed information confidential. | Confidentiality |
| Either party may terminate upon thirty days’ written notice. | Termination |
| This agreement is governed by the laws of New York. | Governing law |
Illustrative rows and categories. Use your own approved labels.
From examples to reviewed results
Prepare
Resolve inconsistent labels and remove unnecessary sensitive information.
Train
Choose a model on your plan and train on representative examples.
Validate
Inspect held-out accuracy by category and the mistakes that remain.
Use and improve
Publish, fill a file or call the API, then review and feed corrections into the next version.
KEEP REVIEW IN THE WORKFLOW
Know where the model stops.
Review clauses covering multiple topics, overlapping labels, and provisions whose meaning depends on other sections. AMT classifies the supplied text; it does not interpret enforceability, extract every clause from a contract, or replace legal review.
Evidence and implementation
Measure before applying labels
Upload labeled clauses, train, and inspect the held-out validation report. Look at errors between related labels rather than relying on the headline score. A classification helps organize review; it does not determine a clause’s legal effect or replace reading the contract.
What the LEDGAR benchmark shows
The LEDGAR article compares four approaches across 100 clause types. Fine-tuning measured 84.9% accuracy in that test. Higher figures discussed with human review are projections that assume flagged mistakes are corrected, not measured autonomous accuracy. Use the full report to understand the data, confusion between labels, and review tradeoffs.
Put it into a review workflow
Fill a file of clause text with suggested categories or call a published model through the API. Have your legal team check consequential or ambiguous results. Correct inconsistent labels and retrain as the taxonomy changes. Use only contract material you are authorized to process.
File fills return category suggestions for your existing import process. API calls return results to your own software; your integration decides how to apply them. Plan allowances and workspace rate limits still apply.
Common questions
Does AMT read the whole contract and extract clauses?
This workflow starts with clause text already prepared by your team or existing extraction process. Supply the text and approved labels in a spreadsheet.
Can it use our own clause taxonomy?
Yes. Train with the categories your team uses. Resolve overlapping definitions and inconsistent labels before training.
Will we get the LEDGAR score?
Not necessarily. LEDGAR measures one public dataset and test split. Validate on your own contracts, labels, and document sources.
Explore the details
Try it on your own examples.
Start with Free for text classification. Compare plans for AI models, Vision, and API access.