Standard System Documentation

Page tree

Attachments: AI

This secondary, AI-focused article explains the differences between the Attachments table in the standard system and the Attachments table in the standard system with AI. Since a KB that has AI enabled requires a more significant build, these details are listed here in order to eliminate confusion in the primary Attachments article.

The Attachments table's structure changes significantly when AI is enabled. The Attachments table contains the tangible files that the extraction model pulls data from, so it is a logical choice for containing a large portion of the AI structure.

Differences

With AI enabled the Attachments table still functions as a way to store files and their qualifying information in record form. With AI enabled, the Attachments table gains many new fields, most of which are populated with data that the AI model extracts from the Attachment record's attached file:

  • Company Fingerprint
  • Contract Type
  • Contract Type AI Document Type
  • AI Closing Date
  • AI Contract Amount
  • AI Contract Description
  • AI Contract Term
  • AI Contract Title
  • AI Doc Type Default Contract Type
  • Document Type
  • Doc Type Specific Fields
  • AI Document Type ID
  • AI Effective Date
  • AI Effective Date Special Format
  • AI End Date
  • AI End Date Special Format
  • AI Governing Law
  • AI Interest Rate
  • AI JSON - Clauses
  • AI JSON - Document Type
  • AI JSON - Key Terms
  • AI Number of Shares
  • AI Party 1 Address
  • AI Party 1 Name
  • AI Party 2 Address
  • AI Party 2 Name
  • AI Payment Term
  • AI Position
  • AI Repayment Term
  • AI Salary
  • AI Service Start Date
  • AI Signature Date
  • AI Termination Notice
  • Button: Extract Clauses
  • Button: Extract Key Terms
  • Button: Predict Document Type
  • Extracted Clauses
  • Extracted Company Fingerprint
  • New Company State Staging
  • Party 1 Name Staging
  • Predicted Document Type
  • Similarity between extracted company and similar company
  • State Name New company ID
  • State Name New company
  • State Abb New company
  • Trigger Clause Extraction
  • Trigger Import Action
  • Trigger Key Terms Extraction
  • Trigger Pushback to Contract

Use Cases

In the standard system with AI, the Attachments table is one of the most important tables. The Attachments table contains the contract files that the extraction model pulls data from, so it is a logical choice for containing a large portion of the AI structure. You can run AI directly from the Attachments table using buttons, as well as integrate the Attachments table with the Agiloft Contract Assistant for Microsoft Word.

Attachment records must have a contract file in the Attached File field. If the Attached File is not in a machine-readable format, it is OCR'd by a create rule, and the resulting OCR File is used for AI. The three main use cases you can initiate from the Attachment table are identifying the contract type, extracting key terms and extracting clauses. These actions are facilitated by the buttons Predict Document Type, Extract Key Terms and Extract Clauses in the Common Area. They are explained in significant detail in the Automation section of this page. The use cases surrounding each button are described in the following sections.

Predict Document Type

The Predict Document Type button allows you to identify the contract type of the document using an AI model. You can also use the Predict Document Type action in a Create rule, so that new Attachment records are classified in the system as soon as they are created. You can classify multiple new documents at once by using a rule like this with a bulk import.

Extract Key Terms

The Extract Key Terms button allows you to extract key terms from the contract file using an AI model. Then, a rule in the Attachments table takes the extracted key terms and inputs it to the mapped fields of the Contract record that is related to the Attachment record. This process effectively allows you to populate the fields of a Contract record simply by having a related record in the Attachments table. You can also use the Extract Key Terms action in a Create rule, so that key terms get extracted from new Attachment records as soon as they are created. You can extract key terms from multiple new contract documents at once by using a rule like this with a bulk import. Adding a new Key Term record to this table doesn't allow the model to recognize or extract that key term, this requires custom model training.

Extract Clauses

The Extract Clauses button allows you to extract clauses from the contract file using an AI model, and then creates individual records for each extracted clause in Contract Clause table. You can then compare these Contract Clauses against the existing, vetted clauses in your Clause Library table to ensure that the clauses in the new contract are legal and do not clash with your company's policies. You can also use the Extract Clauses action in a Create rule, so that clauses are extracted from new Attachment records as soon as they are created. You can extract clauses from multiple new contract documents at once by using a rule like this with a bulk import. Adding a new Clause record to the Clause Library table doesn't allow the model to recognize or extract that clause, this requires custom model training.

Automation

The Attachments table contains three buttons that allow you to extract data directly from a contract file in either the OCR File field or Attached File field of an Attachment record. These buttons contain machine learning actions that use specific AI model configurations to pull the desired data directly from the contract document. After the AI model finds the data, the data is populated as either records or field values within records. These buttons are:

  • Predict Document Type: This button is used to identify the contract type of the contract document. It first ensures that the OCR process is complete if the file requires OCR. Then, CLASSIFICATION-FT-AS AI model is run to determine the contract type of the document. The Document Type field is set using the value returned by the model. Also, if there is a Contract record linked to the Attachment without the contract type selected, that is also updated. Refer AI Document Types for the list of document types supported by the CLASSIFICATION-FT-AS AI model.
  • Extract Key Terms: This button is used to extract key terms from the contract document. It first ensures that the OCR process is complete if the file requires OCR. Then, ATHENA-NER-AS AI model is run to extract key terms. The set of key terms extracted from the document is determined based on the Document Type (either set manually or using Predict Document Type). Extracted key terms are populated in the AI* fields in the Attachments table.
  • Extract Clauses: This button is used to extract clauses from the contract document. It first ensures that the OCR process is complete if the file requires OCR. Then, ATHENA-CE-AS AI model is run to extract clauses. For each individual clause extracted, clause records are created in the Contract Clauses table using an import action.

The Attachments table also contains few new rules that aren't normally found in the standard system. These are:

  • AI/ACA Create Actions: This rule sets File OCR'd to Yes when OCR File is available. Also, it runs PUSHBACK_CLAUSE_UPDATES.jar script action to update Contract Clause records when the Attached File is revised for the first time.
  • Create: AI Extraction - Extract Metadata Triggered on Contract Conversion (Part 1of 4): This rule runs when Trigger Key Terms Extraction is set to Yes. It runs the ATHENA-NER-AS AI model to extract key terms of the contract document. Then, it sets flags to trigger clause extraction and update the contract record with extracted key terms.
  • Edit: AI Extraction - Extract Clauses (Part 2 of 4): This rule runs when Trigger Clause Extraction is set to Yes. It runs ATHENA-CE-AS AI model to extract clauses. Then, it sets a flag to create clause records.
  • Edit: AI Extraction - Create Contract Clauses Records by Import (Part 3 of 4): This rule runs when Trigger Import Action is set to Yes. It runs an import action to create individual clause records in the Contract Clauses table for the clauses extracted in the previous rule.
  • Edit: AI Extraction - Trigger Similarity (Part 4 of 4): This rule runs when Extracted Clauses related table has values and Trigger Import Action is set to Yes. It sets the Extracted Clause Button Clicked flag field to Yes in all the linked Contract Clause records.
  • Edit: Update Contract Fields from Extracted Key Terms: This rule runs when AI JSON - Key Terms field is changed or when the Trigger Pushback to Contract is set to Yes. This rule effectively maps the extracted key terms to the respective fields in the Contract record. If contracting party name is extracted, it is compared with the existing companies using the similarityIndex function to determine if the company already exists in the Companies table. If there is a close match, the contracting party is set to the company found. If not, the extracted data is populated in the Contract record for the user to review and create as a new company.
  • No labels