Artificial intelligence enhanced due diligence
An enhanced due diligence system leverages artificial intelligence to streamline and improve the accuracy of risk assessments for counterparties. The system integrates modules for data extraction, compliance screening, media analysis, and risk evaluation, utilizing large language models and natural language processing to convert unstructured data into structured insights. The system is capable of performing periodic assessments across all entities in a database or on-demand for individual entities, ensuring up-to-date risk assessments. The user interface allows for interaction with the assessment results, providing a comprehensive overview of risk indicators. The system stores risk assessment information providing auditability of prior results.
1 . A computing system comprising:
a hardware computer processor; and
a non-transitory computer readable medium having software instructions stored thereon, the software instructions executable by the hardware computer processor to cause the computing system to execute:
a user interface module configured to determine, based at least on input from a user, a counterparty entity;
a data ingestion module configured to:
ingest unstructured data, including information provided by the user or documents about the counterparty entity;
convert the unstructured data into structured data; and
determine related entities to the counterparty entity;
wherein entities within the unstructured data are resolved via communications with a large language model (LLM) to account for different spellings and nicknames;
a compliance screening module configured to:
identify one or more lists of entities of interest; and
identify, using one or more of matching rules or an LLM, potential instances of the counterparty entity on the one or more lists;
a media screening module configured to:
utilize an LLM agent to generate multiple media screening tasks including separate media screening tasks for each of the counterparty entities and the related entities;
communicate with the LLM regarding the multiple media screening tasks, wherein the LLM is Internet-enabled to allow searching of online media sources;
receive results of search and analysis performed by the LLM regarding the multiple media screening tasks;
communicate with a natural language processor (NLP) to obtain sentiment analysis of results from the LLM; and
generate a media screening overview based on results from the LLM and NLP regarding the multiple media screening tasks; and
a risk assessment module configured to:
for each of a plurality of risk categories, generate a risk indicator based on outputs from one or more of the data ingestion module, the compliance screening module, the media screening module, or organization rules indicating risk tolerances;
wherein the user interface module is further configured to provide a user interface that allows the user to selectively display at least some of the outputs from one or more of the data ingestion module, the compliance screening module, the media screening module, or the risk assessment module.
2 . The computing system of claim 1 , wherein the user interface module is further configured to:
receive feedback from the user regarding accuracy of information in the report; and
initiate updates to one or more models of the risk assessment module based on the user feedback.
3 . The computing system of claim 1 , wherein the multiple media screening tasks include separate tasks for each media type.
4 . The computing system of claim 1 , wherein the multiple media screening tasks include separate tasks for different data ranges or languages.
5 . The computing system of claim 1 , wherein the operations further comprise:
performing the enhanced due diligence process periodically across all entities in a database of the organization or on-demand for an individual entity.
6 . The computing system of claim 1 , wherein the operations further comprise:
storing the outputs from each module for future reference.
7 . The computing system of claim 1 , wherein the operations further comprise:
training one or more models used in the enhanced due diligence process based on training data including snapshots of entities just before they are included on the one or more lists.
8 . The computing system of claim 1 , wherein the counterparty entity is a business entity.
9 . The computing system of claim 1 , wherein the related entities to the counterparty entity include one or more business entity.
10 . A computing system comprising:
a hardware computer processor; and
a non-transitory computer readable medium having software instructions stored thereon, the software instructions executable by the hardware computer processor to cause the computing system to execute:
a user interface module determining a counterparty entity based on user input;
a data ingestion module configured to:
ingest unstructured data, including information provided by the user or documents about the counterparty entity;
convert the unstructured data into structured data; and
determine related entities to the counterparty entity;
wherein entities within the unstructured data are resolved via communications with a large language model (LLM) to account for different spellings and nicknames;
a compliance screening module configured to:
identify one or more lists of entities of interest; and
identify, using one or more of matching rules or an LLM, potential instances of the counterparty entity on the one or more lists;
a media screening module utilizing an LLM agent to search online media sources, generate sentiment analysis, and produce a screening overview, wherein the media screening module generates separate media screening tasks for each of the counterparty entity and the related entities; and
a risk assessment module generating risk indicators for various categories based on data from the other modules and organization rules;
wherein the user interface module is further configured to provide a user interface that allows the user to selectively display at least some of the outputs from one or more of the data ingestion module, the compliance screening module, the media screening module, or the risk assessment module.
11 . The computing system of claim 10 , wherein the data ingestion module utilizes natural language processing techniques to enhance the accuracy of data conversion and entity identification.
12 . The computing system of claim 10 , wherein the compliance screening module accesses real-time updates from global sanction lists to ensure current compliance status.
13 . The computing system of claim 10 , wherein the media screening module categorizes media sources by type and relevance to improve the precision of sentiment analysis.
14 . The computing system of claim 10 , wherein the risk assessment module applies machine learning algorithms to continuously refine risk indicators based on historical data and feedback.
15 . The computing system of claim 10 , wherein the user interface module allows customization of risk assessment parameters based on user-defined criteria.
16 . The computing system of claim 10 , wherein the data ingestion module integrates with external databases to enhance the breadth of data sources available for analysis.
17 . The computing system of claim 10 , wherein the media screening module tracks changes in media sentiment over time and provides trend analysis for the counterparty entity.
18 . A computerized method, performed by a computing system having one or more hardware computer processors and one or more non-transitory computer readable storage device storing software instructions executable by the computing system to perform the computerized method comprising:
determining a counterparty entity based on user input;
ingesting unstructured data, including information provided by the user or documents about the counterparty entity;
converting the unstructured data into structured data;
determining related entities to the counterparty entity, wherein entities within the unstructured data are resolved via communications with a large language model (LLM) to account for different spellings and nicknames;
identifying one or more lists of entities of interest;
identifying, using one or more of matching rules or an LLM, potential instances of the counterparty entity on the one or more lists;
utilizing an LLM agent to search online media sources, generate sentiment analysis, and produce a screening overview, wherein separate media screening tasks are generated for each of the counterparty entity and the related entities;
generating risk indicators for various categories based on data from the other modules and organization rules; and
providing a user interface that allows the user to selectively display at least some of: the structured data, the related entities, the one or more lists of entities of interest, the potential instances of the counterparty entity on the one or more lists, results of the separate media screening tasks, the sentiment analysis; the screening overview; or the risk indicators for the various categories.
19 . The computerized method of claim 18 , further comprising:
receiving feedback from the user regarding accuracy of information in the report; and
initiating updates to one or more models of the risk assessment module based on the user feedback.