Page tree
Skip to end of metadata
Go to start of metadata

Contract Clauses: AI

This table contains fields that store clause details that are extracted by AI.

Clause details extracted by AI are stored in the 'Extracted Clause Text' field. The system will also try to identify matching clauses from the Clause Library using the "SEMANTIC-TEXTUAL-SIMILARITY-AS" model and displays the best match as the 'Most Similar Clause'.

It calculates certain scores, which are represented by fields in the Risk section:

Similarity to Source Clause: This field defines how similar the Extracted Clause is to the Most Similar Clause from the Clause Library table. 0 means the Extracted Clause is not at all similar and 1 means Extracted Clause is exactly same as the clause in the Clause Library table.

Similarity Risk Rating: This is calculated as (1 - Similarity to Source Clause).

Calculated Similarity Risk: This is calculated as the Risk Multiplier of the Most Similar Clause's Clause Type multiplied by the square of Similarity Risk Rating.

Use Case

This table is used to store clauses and clause details that are extracted from contracts and pulled into Agiloft using an AI model called ATHENA-CE. Users don't usually create records in this table; records are generated using a machine learning action.


  • Create: all create rules for contract clauses (Web, API): This is an existing rule that contains additional actions to find the closest matching clauses from the Clause Library.
  • Create/Edit: AI Risk Actions: This rule runs whenever clauses are extracted from a document or are manually tagged, or existing Contract Clauses are modified. IT refreshes the list of matching clauses, sets the best match as Most Similar Clause, and calculates Similarity to Source Clause, Similarity Risk Rating, and Calculated Similarity Risk.

  • No labels