Federal Emergency Management Agency – AI Use Cases

Use Case Name: Planning Assistant for Resilient Communities (PARC) 

Use Case ID: DHS-254 

Use Case Summary: The proposed Generative AI (GenAI) solution, Planning Assistant for Resilient Communities (PARC) will create efficiencies for the hazard mitigation planning process for local governments, including underserved communities. Hazard mitigation plans are not only a foundational step that communities can take to build their resilience but can be lengthy to produce and challenging for communities that lack resources to do so. PARC will specifically support State, Local, Tribal, and Territorial (SLTT) governments’ understanding of how to craft a plan that identifies risks and mitigation strategies as well as generate draft plan elements—from publicly available, well researched sources—that governments could customize to meet their needs. This capability could lead to more communities having the ability to submit grant applications for funding to become more resilient and reduce disaster risks. 

The Beta Release GenAI Plan Generator (OpenAI GPT-4o) is a Large Language Model (LLM) that generates sections of the Hazard Mitigation Plans based on user inputs. It processes prompts and user responses to create detailed, regulatory-compliant Hazard Mitigation Plan sections.  Beta Release PARC Assistant (OpenAI GPT-4o) produces responses through an interactive chat assistant. The Assistant helps planners by answering questions related to the Hazard Mitigation Plan process, offering explanations of policy guidelines, and guiding users through the Federal Emergency Management Agency’s (FEMA’s) regulatory requirements.

Use Case Topic Area: Emergency Management 

Deployment Status: Pre-deployment (Acquisition and/or Development) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR /FC)? No 

Use Case Name: Hazard Mitigation Assistance Chatbot 

Use Case ID: DHS-2439 

Use Case Summary: The Hazard Mitigation Assistance (HMA) Chatbot, will be developed for Internal Stakeholders Navigating HMA Grant Applications with Benefit-Cost Analysis (BCA) Assistance, Project Scoping, and Feasibility Support to serve as a centralized, intuitive user interface for staff across all Federal Emergency Management Agency (FEMA) regions and headquarters. The AI-powered chatbot will utilizing Natural Language Processing (NLP) to provide tailored, context-specific guidance to assist FEMA HMA staff, including new hires, who face significant challenges in accessing and processing vast amounts of information related to applications and grants. Benefits include accelerating onboarding of new hires, reducing the time to full productivity, ensuring compliance and accountability through transparency and accuracy, reducing errors, enhancing decision-making by providing access to precise citations and real-time data, and improving the quality of decisions which benefit program outcomes. The chatbot aims to enhance overall efficiency and service delivery, aligning with FEMA's goal of improving program delivery and serving the whole community. 

The chatbot will provide accurate, uniform answers with citations, ensuring everyone is aligned and reducing the risk of misinterpretation of publicly available HMA data sourced from FEMA.gov. The chatbot will provide clear, plain-language explanation, highlighting eligibility criteria, funding priorities, and application processes for each HMA program. The chatbot will quickly retrieve relevant HMA policy documents, provides precise citations, and summarizes key points. The chatbot will compile publicly available HMA data sourced from FEMA.gov, perform analyses, and present it in a clear report with visualizations. 

Use Case Topic Area: Emergency Management 

Deployment Status: Pre-deployment (Acquisition and/or Development)  

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Recovery and Resilience Resource 

Use Case ID: DHS-2440 

Use Case Summary: The Recovery and Resilience Resource (RRR) Portal aims to simplify the process of finding and using disaster recovery and resilience resources and information for State, Local, Tribal, and Territorial (SLTT) decision-makers as well as the communities they serve. The RRR Portal brings together technical, financial, and information assistance into one easy-to-use platform to reduce the time it takes to navigate the wide range of available resources. The non-AI layers (currently funded) will provide an intuitive user interface, a business intelligence reporting tool, and an expansive data library. The AI-layer (currently unfunded) would include a smart matching wizard functionality via AI Large Language Model (LLM) to connect SLTT partners with the resources that meet their unique needs. 

Increasing access and improving navigation of federal resources will benefit a wide array of stakeholders across all sectors, including private and public. The primary stakeholders are SLTTs that are looking to support recovery and/or resilience in communities. This includes 50 states, Washington D.C., 5 U.S. territories, about 80,000 local governments (including special districts), 574 federally recognized tribal governments, and their partners. Additional stakeholders include the 25+ federal departments and agencies that are members of the Mitigation Framework Leadership Group (MitFLG), who have equity in the availability of comprehensive resources, best practices, or tools for SLTTs to recover from or increase resilience to disasters. 

The smart matching wizard will be an AI LLM that analyzes user search behavior and matches recommended resources from across the expansive data library that will include resources from federal, state, local, nonprofit sectors. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Pre-deployment (Initiation) 

Safety- and/or rights-impacting? No, use case is too new to fully assess impacts; will be reassessed before end of initiation stage. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Digital Processing Procedure Manual (D-PPM) 

Use Case ID: DHS-2442 

Use Case Summary: The Digital Processing Procedures Manual (D-PPM) technology is an AI-based search that will be implemented in call centers and remote field services locations to enhance agent support, encourage proper utilization of guidance and procedures, and ensure a consistent, empathic survivor experience. Leveraging this intelligent retrieval technology does not rely on extensive application changes and offers substantial benefits in providing round-the-clock support while prioritizing complex inquiries. This initiative aims to enhance survivor satisfaction and call center efficiency by reducing call handle times and maintaining or improving answer quality.  The current applications for handling agent inquiries for processing applications are resource-intensive, leading to high operational costs and inefficiencies, especially during peak periods immediately following presidential-declared disasters. Currently, FEMA has a repository on SharePoint called the Processing Procedures Manual (PPM) that houses more than 70 documents that make up over 2,000 pages worth of content and guidance. Agents search this content and guidance with find function and keyword search, first having to find the right document, and then using keyword-based search and find functionality, negatively impacting efficiency and quality and leading to increases in escalations to higher tiers of agents/support who are no longer freed up to work on more complex situations.  Implementation of a 24/7 digital, interactive application will expedite guidance comprehension and discovery to provide better support to disaster survivors during active disasters. The system would use AI and search capabilities to intelligently analyze FEMA policies and guidance and assist the agents through the assistance review and award process.  

The AI will enable agents to retrieve information quickly, reducing the overall time spent on each call.  By providing more accurate and relevant search results, the AI will help agents deliver consistent, correct responses.  Reducing the time spent per call will improve operational efficiency, leading to potential cost savings.  By empowering agents with the correct information, the tool will lower the frequency of escalations to supervisors, allowing them to focus on more complex cases.  The time saved through the AI tool will allow agents to handle more complex queries, ultimately improving service quality and survivor satisfaction.  As the AI tool is rolled out, it will provide a scalable solution that can grow with the agency’s needs during large-scale disaster responses. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Pre-deployment (Initiation)  

Safety- and/or rights-impacting? No, use case is too new to fully assess impacts; will be reassessed before end of initiation stage. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Incident Management Workforce Deployment Model (depmod) 

Use Case ID: DHS-248 

Use Case Summary: The Incident Management Workforce Deployment Model helps predict how FEMA's incident management team might respond to disasters. It uses historical data to plan staffing needs but does not directly decide who gets staffed.  The model is used to make predictions about how the incident management (IM) workforce could respond to the Stafford Act incidents. A key area of application is the setting of IM workforce staffing levels (i.e., force structure). The model was constructed in R and delivered as an R package. The model uses various kinds of statistical inference and machine learning methods (i.e., multinomial regression, Bayesian multi-level modeling, etc.).  This use case leverages historic data on the deployment of FEMA personnel to anticipate requirements for disaster staffing and supporting assessments for overall readiness. It does not leverage demographic data, nor is it used to directly inform staffing decisions.  It is purely used for planning and assessment purposes.  

The model is able to extract patterns from big data, combine these patterns in a way that mirrors a real-world process, and simulate likely outcomes to yield predictions in the form of relatable, mission-centric metrics. 

The system primarily outputs predictions that can then (optionally) be summarized and communicated as recommendations. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Individual Assistance (IA) & Public Assistance (PA) Projections 

Use Case ID: DHS-251 

Use Case Summary: Individual Assistance (IA) and Public Assistance (PA) Projections predicts recovery program quantities of interest using supervised learning models. These models include, but not limited to, predicting the number of households that will register for the Individuals and Households Program (IHP), the number of households that will require direct housing assistance, the number of temporary transitional housing units needed, the number of applicants for Public Assistance, the number of PA projects that applicants will submit, the number of sites that will need to be inspected per PA project, and the cost of delivering assistance.  Supervised learning models include sample statistics, generalized linear models, decision trees, and deep neural networks. 

The purpose of these supervised learning models is for informational purposes, to produce predictions for to-be-determined quantities of interest and for decisional purposes that may include implicit use (e.g., seeing a large number of households requiring assistance may motivate FEMA Chief Financial Officer (CFO) to consider requesting additional funding for the Disaster Relief Fund) or explicit use (e.g., the number of estimated housing units required may be used to determine the number of housing units to order just-in-time). Anticipated benefits of using supervised learning models include increased predictive performance, timely predictions, quantification of uncertainty, and rigor. 

These supervised learning models produce predictions, not recommendations or decisions (though they can inform human users in making recommendations and decisions). At minimum, these models will produce point predictions for the different quantities of interest for disaster declarations. Additionally, they may produce prediction intervals or predictive distributions as feasible and appropriate for the given prediction problem. Often these outputs will be shared via business intelligence tools (e.g., Tableau or PowerBI) for wide internal FEMA use. Some predictions may be shared with a more restricted audience through simpler means (e.g., an excel workbook) as appropriate. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Geospatial Damage Assessments 

Use Case ID: DHS-346 

Use Case Summary:  The Response Geospatial Office (RGO) is exploring the use of AI to assist in the prioritization of structural and debris assessments. RGO reviews satellite, aerial, and radar imagery to expedite damage assessments in the aftermath of a disaster. RGO is utilizing several AI techniques, including computer vision, machine learning, and deep learning, to help prioritize imagery exploitation by identifying areas likely having damage or debris. Areas indicated as likely having damage or debris are prioritized for imagery exploitation by analysts, reducing the time required for visual damage assessments. FEMA will integrate these automated tools for imagery-based damage assessments into RGO processes and systems to inform FEMA geospatial analysts. FEMA geospatial analysts will review output and develop recommendations, so FEMA can quickly detect and characterize damaged and undamaged buildings, identify concentrations of damage, detect debris, and support rapid response and recovery activities. AI models generate points or polygons over areas where damage or debris is likely. This output is a recommendation meant to trigger investigation by human analysts via traditional damage assessments. In large-scale incidents, imagery collection can be comprised of hundreds of thousands of miles of land and millions of structures. AI automation allows for the prioritization of imagery exploitation where damage is likely to have occurred, reducing the time required to identify impacts and summarize damage sustained. 

Use Case Topic Area: Emergency Management 

Deployment Status: Deployed (Implementation and Assessment)  

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: OCFO Response Augmentation Suite 

Use Case ID: DHS-2296 

Use Case Summary: FEMA Office of the Chief Financial Officer Generative Pre-trained Transformer (OCFO GPT), Travel Policy GPT and Fiscal Policy GPT are internal Generative AI (GenAI) tools designed to support the FEMA workforce by generating initial responses to various queries. These tools leverage relevant public and internal documents to draft preliminary responses, which are then refined prior to formal submission. They assist in the data gathering stage, but do not replace the critical review and judgement of FEMA analysts and leadership.  FEMA OCFO GPT generates initial responses to questions for the record, leveraging public and internal documents, and provides a preliminary response to the Program Office to use in their formal response to the request. It reduces the data gathering stage, saving analysts 80% of the initial effort.  Travel Policy GPT generates initial responses to questions regarding FEMA/DHS Travel Policy, including the JTR, and provides a preliminary response to the travel specialist to use in their formal response to the queries.  It improves response times, saving users 80-90% of the time compared to regular engagement with the Travel Service Center.  Fiscal Policy GPT provides preliminary responses to questions regarding FEMA/DHS Fiscal Policy and will generate a draft response with references to assist FEMA internal workforce in compliance with established policy.  It saves users 80-90% of the time compared to regular engagement with DHS /FEMA OCFO policy and speeds up resolution times.  These tools do not replace analysts’ work or leadership review, but enhance efficiency in data gathering and preliminary response stages. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: OCFO Code Assist GPT 

Use Case ID: DHS-2441 

Use Case Summary: Code Assist Generative Pre-trained Transformer (GPT) is an internal facing Generative AI (GenAI) tool to augment the FEMA workforce in generating and troubleshooting existing queries in established query languages (e.g., SQL, Java, COBOL). Users enter the language they are querying, and the Code Assist GPT then provides a proposed query based on the elements provided.  If the query is unsuccessful, the tool maintains the session, allowing users to prompt for enhancements until expected results are achieved.  At the end of the session, all prompts and queries are removed, and no data is stored outside of the active session. 

The tool provides improves query generation and rapid iteration, saving users 80-90% of the time compared to custom query development.  It also supports various computer languages to assist the data analytics community. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Implementation and Assessment) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: FEMA OCFO GPT  

Use Case ID: DHS-250 

Use Case Summary: FEMA Office of the Chief Financial Officer Generative Pre-trained Transformer (OCFO GPT) is an internal facing Generative AI (GenAI) tool to augment the FEMA workforce in generating initial responses to questions for the record and providing preliminary responses to questions for the record and providing preliminary responses for the Program Office. The tool leverages public facing and internal deliberative documents (e.g., hearing testimony, answers sent to previous questions) to assist in answering questions the Agency receives. It generates draft responses that are then refined before formal submission. This tool is used in the data gathering stage and does not replace the work of analysts or the review by Agency leadership.  It reduces the analyst initial level of effort versus individual research. 

Deployment Status: Inactive (consolidated with another use case).  This use case was consolidated under OCFO Response Augmentation Suite (DHS-2296).