Click to view related learning resources:
Help Center
-
AI Clauses and AI Custom Clauses Overview: This article providesan overview of AI clauses and custom AI clauses. This includes a list of best practices when managing clauses.
- AI Clauses Overview: This article provides an overview of AI clauses. This includes a list of preconfigured clauses and how to manage them.
- AI Custom Clauses Overview: This article provides an overview of AI custom clauses. This includes how to manage them and a list of frequently asked questions.
Academy
- Standardize Contract Negotiations with Ironclad AI: Learn how to expedite your contract review processes and work more efficiently with Ironclad AI features.
This article will walk you through how to train Ironclad AI clauses and AI custom clauses.
You can train both Ironclad out-of-the-box AI clauses and AI custom clauses by providing a few examples of clause text. You can start with 1-2 examples, adding and deleting examples at any point in time based on your business’ needs.
Clause Type | Description |
---|---|
AI Custom Clauses
|
|
Ironclad AI Clause
|
|
Train an Ironclad AI Clause and AI Custom Clause
- Click on your profile icon located in the top right corner of Ironclad and navigate to Company Settings > AI Clauses.
- If you want to create an AI custom clause to train:
- Click + Add New to create a new Ironclad AI clause.
- Enter a name and a description of the clause type.
- If you want to train an existing AI clause, click on the clause you want to train.
- Add an example to train the AI to detect clauses in documents. Based on the examples provided, the AI can then learn to detect clauses in AI Playbooks and Smart Import. You can add up to 20 examples per clause.
-
As you add an example, you can see what the example’s Training Impact is. The training impact will give you an idea on how similar your example is to another example. Based on the example provided, the training impact will display as Weak, Medium, or Strong.
- Weak doesn’t mean the example won’t be helpful to train the AI, but rather that it’s similar to existing examples.
- We generate training impact by using a clause embedding model to define vector representations of each clause. While these vectors aren't directly interpretable by humans, they effectively capture the essence of the clauses. For example, a vector might indicate how closely a clause aligns with the characteristics of a confidentiality clause.
- After you add examples to your AI custom clause, underneath Detection Settings, you can use Similarity Settings to control if and when the clause will be detected in AI Playbooks and Smart Import documents. When using AI Playbooks or Smart Import, your AI custom clauses will detect clauses in the document. They will detect clauses that meet your defined detection criteria. There are three similarity setting options:
-
-
-
Exact match: AI will only detect clauses in documents that are exactly the same as the examples you have provided.
NOTE
The first four words will be ignored, and if there are new lines added in the document text, the AI will not be able to detect it correctly. - Highly Similar: AI will detect clauses in documents that look highly similar to the examples you have provided to train the AI custom clause. We recommend this option for most use cases.
-
Somewhat similar: AI will detect clauses in documents that look somewhat similar to the examples you have provided to train the AI custom clause. By choosing this option, AI will detect the most results compared to the other options.
-
Exact match: AI will only detect clauses in documents that are exactly the same as the examples you have provided.
-
-