AI Solution for TotalAgility
AI Solution for TotalAgility Overview
Features
Our AI-enabled document analysis solution allows gathering of relevant insights from unstructured or varied-structure documents.
Examples of some use cases and outputs for this solution:
- Asking AI (instead of a human reviewer) if an insurance policy includes certain coverages
- Asking AI to confirm whether a contract has a certain attributes like whether or not there's an early termination fee
- Retrieving information from documents that have significantly different layouts where traditional capture is not possible or not practical
- Summarizing or finding relevant data in long documents that are difficult or time-consuming for human users to process
Benefits
AI model selection on a per-question basis supports using more powerful models where needed or faster and less expensive models where feasible.
Searchable vector store implementation allows for sending only specific relevant portions of a larger document to the LLM which supports more relevant results, faster performance, and lower costs.
Technical Details
Inputs
Allows for a configurable list of questions to ask the AI service per document type, with configuration options (adjustable per question) including:
- OpenAI LLM model to use
- Options to return either a plain text answer or a JSON list for cases where a list or table of information is needed
- Parameters for citation coordinate (ie. location information for where answers were found) handling
- Vector store query settings
Utilizes embedding of document text in a searchable vector store. When searched, this provides only the most relevant context to the LLM for processing.
Provides document page and coordinates of a citation for where in the document the provided answer was found.
Provides an explanation from the AI engine of why the AI determined the given answer to be correct, allowing for additional verification of and confidence in results.
Geographic Availability
Additional Information
AiGS's solution can be demonstrated and implemented for both new and existing Tungsten TotalAgility systems.
