AI Use Cases in Department of State

Introduction

The Department of State (DoS) can leverage AI to enhance its operations across various functions, from procurement and logistics to foreign assistance and conflict forecasting. By integrating advanced AI technologies, the DoS can improve efficiency, accuracy, and decision-making in its diverse activities. The following use cases illustrate how AI can be utilized to advance the Department’s capabilities and support its mission.

Use Cases

  • 1. Federal Procurement Data System (FPDS) Auto-Populate Bot

    This project involves the development of a bot to automate data entry in the Federal Procurement Data System (FPDS), reducing the workload on procurement staff and improving compliance with the DATA Act. The bot updates approximately 300 FPDS awards per week. Additionally, bots have been developed to automate closeout reminders for federal assistance grants, receiving report validation, and customer service inbox monitoring, enhancing efficiency and accuracy in these processes.

  • 2. Product Service Code Automation ML Model

    This project involves a machine learning model that scans unstructured procurement data entered by users, such as Requisition Titles and Line Descriptions, to automatically detect and categorize the types of commodities and services being purchased. This automation improves the accuracy and efficiency of procurement categorization.

  • 3. Tailored Integration Logistics Management System (ILMS) User Analytics

    This project aims to use transactional data from the Integrated Logistics Management System (ILMS) to develop tailored user experiences and analytics. By analyzing real system actions and clicks, the project seeks to extract meaningful information to simplify user interactions with the system and reduce the time required to complete daily tasks, enhancing overall user efficiency and satisfaction.

  • 4. Supply Chain Fraud and Risk Models

    This project aims to enhance risk analytics by developing AI/ML models to detect unusual activities within the Integrated Logistics Management System (ILMS) that could indicate fraud or misconduct. The models will build on existing risk models and target key supply chain functions like Asset Management, Procure-to-Pay, and Fleet Management, improving the detection and prevention of fraudulent activities.

  • 5. Within Grade Increase Automation

    This project uses an NLP model along with Intelligent Character Recognition (ICR) to identify and extract data from the JF-62 form for within-grade increase payroll actions. Robotic Process Automation (RPA) is then employed to validate the data against existing reports and create a formatted file for approval and processing, streamlining the payroll action process.

  • 6. Verified Imagery Pilot Project

    This pilot project tested the use of Sealr, a technology service that verifies the delivery of foreign assistance to conflict-affected areas where neither the U.S. Department of State nor its partners can go. Sealr uses blockchain encryption to secure smartphone photographs from tampering and AI to detect spoofing. The pilot showed that such technology can enhance remote monitoring of foreign assistance in dangerous or inaccessible areas.

  • 7. Conflict Forecasting

    This project involves developing models to forecast conflict and instability using open-source political, social, and economic data. The models predict outcomes like interstate war, mass mobilization, and mass killings, using AI techniques such as tree-based methods, neural networks, and clustering. These models aim to improve the prediction and understanding of conflict dynamics.

  • 8. Automated Burning Detection

    This project uses AI and machine learning to scan moderate resolution commercial satellite imagery daily, identifying anomalies in the near-infrared band. The goal is to detect burning activities, enhancing monitoring and response capabilities in affected areas.

  • 9. Automated Damage Assessments

    This project employs AI and machine learning to analyze moderate and high-resolution commercial satellite imagery, documenting war crimes and other abuses in Ukraine. It includes automated damage assessments of various buildings, such as critical infrastructure, hospitals, schools, and crop storage facilities, providing valuable data for accountability and recovery efforts.

  • 10. NLP for Foreign Assistance Appropriations Analysis

    This project uses natural language processing (NLP) to automate the extraction of earmarks and directives from the annual appropriations bill for the Foreign Assistance Appropriations Analysis. Previously, this task was done manually, but NLP streamlines the process, making it faster and more efficient.

  • 11. eRecords M/L Metadata Enrichment

    This project uses machine learning to enhance the metadata of the Department’s central eRecords archive. The models perform tasks such as entity extraction, sentiment analysis, classification, and document type identification, improving the discovery and review of records.

  • 12. Facebook Ad Test Optimization System

    This system collects and analyzes media data from various sources to provide a comprehensive and current global view of media coverage. It helps optimize Facebook ad tests by leveraging this extensive media data.

  • 13. Global Audience Segmentation Framework

    This prototype system gathers and analyzes daily media clips from around 70 Embassy Public Affairs Sections. It helps segment global audiences by providing detailed insights into media coverage and public opinion.

  • 14. AI Capabilities Embedded in SMART

    This project integrates AI models into the SMART system on OpenNet to enhance cable processing. The models perform entity extraction, sentiment analysis, keyword extraction, and historical data analysis, helping users compose cables more efficiently by recommending addressees and passlines.

  • 15. NLP to pull key information from unstructured text

    This project uses natural language processing (NLP) to extract key information, like country names and agreement dates, from unstructured PDF documents. This automation simplifies the process of retrieving important data from lengthy documents.

  • 16. K-Means clustering into tiers

    This project uses k-means clustering to categorize countries into tiers based on data collected from open sources and bureau data. This classification helps in better understanding and managing country-specific information.

  • 17. Optical Character Recognition – text extraction

    This project employs optical character recognition (OCR) using standard Python libraries to extract text from images. The inputs for this process include data collected from websites, making it easier to digitize and analyze textual information from images.

  • 18. Topic Modeling

    This project uses topic modeling to group text into themes based on word frequency. It has been applied to digital media articles and social media posts using Python libraries, helping to identify prevalent topics and trends in large text datasets.

  • 19. Forecasting

    This project employs statistical models to predict future outcomes, such as COVID-19 case numbers and violent events based on tweet analysis. These forecasts help in planning and response efforts by providing insights into potential future scenarios.

  • 20. Deepfake Detector

    This project uses a deep learning model to detect deepfakes by analyzing images of faces. It classifies images as real or fake, helping to identify synthetically generated faces created using techniques like Generative Adversarial Networks (GANs).

  • 21. SentiBERTIQ

    This project uses a sentiment analysis model, SentiBERTIQ, to identify and extract subjective information from text. Trained on 2.2 million tweets in multiple languages, it assigns sentiment scores to text documents and outputs the results in a CSV file for review, aiding in understanding public sentiment across different languages.

  • 22. TOPIQ

    This project involves the TOPIQ tool, which uses Latent Dirichlet Allocation (LDA) to automatically classify text into topics. It helps analysts review and interpret large collections of documents by identifying and categorizing the main topics within them.

  • 23. Text Similarity

    This project uses a text similarity tool to identify identical or nearly identical texts by calculating cosine similarity. Texts with high similarity are grouped together, making it easier for analysts to review and compare related documents.

  • 24. Image Clustering

    This project uses a pretrained deep learning model to create image embeddings and then applies hierarchical clustering to identify similar images. This helps in organizing and analyzing large collections of images by grouping visually similar ones together.

  • 25. Louvain Community Detection

    This project uses the Louvain Community Detection algorithm to analyze social networks by clustering nodes into communities. It groups similar nodes together, helping to identify and understand the structure and dynamics of social networks.

Conclusion

By integrating AI into its operations, the Department of State can significantly enhance its capabilities in areas such as procurement, logistics, and conflict forecasting. These use cases demonstrate the potential for AI to improve efficiency, accuracy, and decision-making processes across various functions, supporting the Department's mission more effectively and addressing complex challenges with advanced technological solutions.

Discuss a Use Case

Fill in your details & we will get back to you shortly.