DENVER — The National Geospatial-Intelligence Agency is expanding its use of artificial intelligence to process growing amounts of geospatial data, even as officials warn that expectations for continuous, real-time insights remain out of reach.
The rapid adoption of AI-driven analytics is raising expectations that geospatial intelligence can provide near-constant awareness and reduce the time from data collection to analysis. But NGA deputy director Brett Markham says that perception outweighs the reality.
“There are a certain number of people in this world who want to know everything about everything 24 hours a day, 365 days a year,” Markham said in his keynote speech at the GEOINT Symposium on May 3. “Some people think we have that ability today, and I hope that’s true.”
NGA is a U.S. intelligence agency within the Department of Defense that collects, analyzes, and disseminates geospatial intelligence (information derived from satellite and other location-based data) in support of military operations, national security, and disaster response.
AI models now automate much of the image analysis, detect objects, and detect anomalies at scale. However, algorithms are limited in their ability to interpret context like a human analyst.
Markham said this gap is impacting how NGA approaches AI investments. Rather than treating automation as a solution in itself, the agency is using automation to reduce wait times for intelligence analysts and narrow uncertainty.
“We want to be able to automate certain workflows that allow analysts to get information in minutes instead of hours, and apply artificial intelligence to quickly provide information to decision makers,” Markham said. “The signals of demand for more accurate and timely information will continue to grow,” he said, pointing to heightened expectations across military operations in the space, air, sea and land domains.
The changes come as NGA faces an explosion of data from satellites and other sensors, forcing changes in how intelligence is processed. Analysts are increasingly relying on AI “agents” to identify objects and surface anomalous activity, allowing humans to focus on interpretation rather than initial detection.
A key area of development is the use of multimodal AI models. These are systems that combine multiple data types into a single analysis pipeline. Geospatial intelligence includes optical satellite imagery, synthetic aperture radar, infrared data, and non-imaging sources such as text reports and signal metadata.
This approach is designed to improve analyst productivity. While optical images can be obscured by weather or darkness, radar data produces different representations of the same scene. By integrating inputs, multimodal systems can maintain analysis continuity even when one source degrades.
Although much of NGA’s AI work is done on classified systems, NGA is increasingly relying on commercial technology. “We don’t have the time or expertise to build cutting-edge AI models from scratch,” Markham said, referring to cutting-edge systems developed by a small group of companies.
The Department of Defense is moving toward formalizing these relationships. On May 1, the company announced agreements with several leading artificial intelligence companies, including OpenAI, Google, Nvidia, Microsoft, and Amazon Web Services, to deploy AI capabilities to the Department of Defense’s sensitive networks.
Efforts are also underway within NGA to accelerate the acceptance of AI tools. The agency launched the Computer Vision Model Certification Campaign, inviting companies to participate in a 90-day process to validate algorithms for national security applications. Because these systems are considered vetted for operational use, companies with certified models may have an advantage when competing for government contracts.
These computer vision models are trained on satellite, aerial, and drone imagery and are designed to do more than identify objects. Analyzing images associated with specific locations over time helps analysts infer activity and track changes, forming the backbone of systems such as the military’s Maven smart system, which is now widely used across military and intelligence agencies.
