Potential use cases for generative AI in public health can be categorized into four key domains.
Click on each use case to reveal more
Service delivery and operations
Resilience, preparedness, and outbreak response
Product R&D
Foundations for public health action
¹Maternal, newborn, and child health.
Accelerate grant writing by improving speed and quality of research, drafting, and tailoring of grant applications Aggregate nontraditional sources of qualitative data (eg, social media and speech-to-text feedback surveys) to identify issues in near real time (eg, shortages of malaria bed nets) Capture insights from unstructured data to inform models of risk scenarios and mitigation strategies, structurally improving investor confidence in global health innovation opportunities
Supply chain and financing
Generate campaign communications (eg, images and text) tailored to local language and context Provide enhanced telemedicine and point-of-care support through gen AI–enhanced chatbots to improve healthcare in rural areas (eg, antenatal care and MNCH¹ outcomes) Generate personalized treatment plans and patient guidance based on individual patient data, delivered in a format tailored to lifestyle and literacy level
Develop individualized training content and simulations for physicians, nurses, and healthcare workers tailored to needs (eg, region, area of specialty, and demographics served) Synthesize care coordination notes across patient’s care team (eg, nurses, primary care physicians, doulas, and midwives involved in antenatal care) Enhance incomplete or low-quality patient data (eg, build a visual of patient tissue or organs from an MRI scan) to aid in the diagnostic process
Frontline healthcare worker enablement
Community and patient engagement
McKinsey & Company
Summarize trends and latest developments from scientific literature search to improve existing models and shape new approaches to various conditions Automatically generate content for response frameworks and scenarios for table-top exercises to test and improve readiness Synthesize learnings from after-action review reports to generate best practices and inform response frameworks for future public health threats
Predictions and preparations for threats
Address operational challenges in the first-responder community (eg, assist emergency responders in quickly generating tailored communications in different languages from existing collateral) to improve timeliness and effectiveness of aid Generate and enhance tailored health guidance that health officials can review and disseminate to the public during an outbreak Improve decision support across the public health ecosystem through answering questions about and extracting content from daily after-action review reports, medical knowledge, electronic medical records, and other knowledge sources
Leverage image processing across multiple data sources (eg, thermal mapping, satellite images) to generate insights and flag aberrations Identify early-warning factors or signals (eg, synthesize emergency department visit notes, triage notes, and signals from social media for syndromic surveillance) to enhance syndromic surveillance and ensure rapid awareness of potential threats
Detection and monitoring of potential threats
Investigation of and response to threats
²Lower- and middle-income countries. ³Clinical study reports. ⁴Medical legal reviews. ⁵Investigational new drug. ⁶New drug application.
Synthesize insights from translational research for disease elimination (eg, monitoring uptake of a disease preventive therapy across different LMIC² contexts or across organizations) Understand and summarize biomedical research papers to identify papers that are most relevant for global health Enhance indication finding and drug repurposing through representational learning from patient histories
Discovery and research
Generate documentation needed for regulatory approval using natural language processing and through analyzing prior submissions and synthesizing real-time drug development or clinical trial data (eg, auto-drafting CSRs,³ MLRs,⁴ IND,⁵ and NDA⁶ submissions)
Use deep learning to design protein molecules and advance new protein-based vaccines for diseases (eg, leveraging data from vaccine or drug libraries and medical legal review) Integrate, clean, and synthesize data across trial sites to improve data quality and expedite analysis or summaries for dossier submission Automatically generate case report forms based on digital protocol, and run AI-based automated data quality checks and cleaning
Development
Regulatory approval
Create synthetic data sets to test and develop solutions without risking privacy breaches or providing personally identifiable information Automate administrative tasks to reduce manual burden (eg, HR chatbot for self-service, processing large amounts of unstructured data) Correct and troubleshoot code to accelerate engineering productivity by more than 50 percent; accelerate pair programming with generative AI “programmer” (eg, Codex, AlphaCode, Codegen, and Copilot)
Data and technology enablers
Analyze and synthesize policy documents across countries to determine best practices and tailor global health interventions to regional contexts Simulate various scenarios and predict the impact of potential policies and standards to aid decision making Generate draft data policies and standards (eg, by synthesizing trends and impact of prior or other policies and standards across different geographies)
Enable engineering talent to search for creative solutions, decipher unfamiliar syntax, and find the correct algorithm when unsure of how to proceed Create personalized learning journeys and recommend learning resources based on current skills and future goals
Talent enablers
Policies and standards
