In his 1966 book Men, Machines, and Modern Times, Etling E. Morison offered what would become a classic study of how people and organizations learn, or fail to learn, to live with innovation. One of my favorite vignettes is his account of an early World War II artilleryman acting as a “horse holder,” standing as if to calm the team of horses during firing, even though the gun was now pulled by a truck. The absurdity of that remnant of battlefields past comes to mind as leaders envisage how geospatial intelligence is produced—and leveraged—in the age of autonomy.
Over the past forty-five years, I have closely followed how innovations in geospatial intelligence, or GEOINT, and its predecessor fields have served the warfighter. The field has changed over the years to follow technology trends and warfighter needs, but the current pace of change is faster, broader, and more consequential. To keep pace, a new evolution—what I call autonomous GEOINT—must emerge as the next, necessary step to continue supporting warfighting on the rapidly emerging autonomous battlefield.
Building upon the definition of GEOINT in 10 US Code § 467(5), today GEOINT provides warfighters with the locations of US forces, coalition partners, adversaries, and noncombatants. It helps predict movement by accounting for natural and human-made obstacles. The advantage GEOINT affords battlefield leaders goes beyond describing what, where, and when; it also helps to address how and why.
My experience with the transition from the Defense Mapping Agency to the National Geospatial-Intelligence Agency and the Army’s shift from topographic engineering to geospatial engineering suggests that technology is often not the hardest problem. While both of these transitions were compelled by significant technological and warfighting changes, the harder problem was how people accepted the changes, adapted to new organizations, and created doctrine to make new capabilities useful. These professional communities metaphorically had their horse holders standing in place while the guns were hitched to more capable trucks.
I fear that today’s GEOINT community may be repeating that pattern of horse holders standing in place, not fully grasping how lessons being learned from the conflicts in Ukraine, Gaza, and Iran, and from the portion of space defined as low earth orbit (LEO), are reshaping GEOINT. The trends are clear: drones, autonomy, electronic warfare, low-cost mass, underground protection, efforts to deny systems in LEO, kill webs, deception, and human-machine teaming. What I have found difficult is explaining what those trends mean, practically, for GEOINT.
That changed when I read David Petraeus and Isaac Flanagan’s Foreign Affairs article, “The Autonomous Battlefield.” They argue that autonomous warfare is emerging in Ukraine and the Middle East. As electronic warfare breaks the communications links necessary for human control, missions and operational plans are being executed by autonomous drones. Their writing helped me see how autonomous warfare will reshape GEOINT missions, products, workflows, human roles, and operational tempo. GEOINT’s new task may be to encode geographic knowledge for autonomous systems. This is autonomous GEOINT—geospatial intelligence designed to support autonomous systems.
What Current Conflicts Show
Expanding upon Petraeus and Flanagan’s teachings beyond autonomous warfare, the conflicts in Ukraine, Gaza, and Iran, along with the emerging contest in LEO, reveal other critical lessons.
The battlefield is more transparent as commercial satellites, drones, sensors, connected devices, and AI-based analytics make it harder to hide, move, resupply, and maintain infrastructure. Time and geographic space are compressing. Sensors, orbital communication networks, and machine processing reduce latency and let tactical actors observe, track, and influence targets across distances once associated with the operational and strategic levels of war. Decisions must increasingly be made quickly in near real time, even as complex considerations of terrain, weather, logistics, physics, risk, and human judgment still matter. Cheap mass is changing force structures. Low-cost, attritable systems can saturate defenses, deplete interceptors, blind sensors, and attack high-value targets in waves. Exquisite systems still matter, but cheap mass systems change the cost curve and force new habits.
Rapid adaptation is also becoming a combat function. Success increasingly depends on turning experience into software updates, training data for AI systems, hardware changes, production shifts, and developing new tactics faster than the opponent can copy or counter them. Electronic warfare, navigation warfare, and signature management are moving to the center of the fight, requiring GEOINT to account for jamming, spoofing, degraded connectivity, electromagnetic emissions, relay nodes, drone operator locations, and disrupted positioning, navigation, and timing (PNT). Concealment and deception are returning as core operational functions. As transparency expands, survival depends on camouflage, concealment, decoys, hardening, underground facilities, signature management, and electronic terrain modification. Ukraine, Gaza, and Iran show that underground protection is not always sanctuary; it works best when combined with dispersion, redundancy, deception, and protected logistics. Even so, the X-factor remains human and institutional ingenuity—the intangible quality that often determines whether something becomes truly successful. Machines may accelerate the fight, but humans still have to think.
Implications for GEOINT
Petraeus and Flanagan make two key points about future warfare. First, autonomous systems are here and will increasingly be organized into formations that move and behave as military units. And second, they need to execute a mission based on the commander’s intent and concept of operations even when completely disconnected from human control.
A human platoon, company, battalion, or brigade uses specific geographic concepts and knowledge, which include boundaries, routes, phase lines, fire support coordination measures, checkpoints, logistics, protected areas, and rules of engagement. Autonomous formations will need the same geographic knowledge in machine-readable form. This encoded knowledge will help the machine to predict, solve problems, recognize patterns, and adjust behavior in the fog of war.
At the platoon level, GEOINT support for autonomous systems may focus on microterrain, obstacles, routes, visibility, building access, electronic risk, and friendly-force deconfliction. A small autonomous formation may need a narrow operating box, a route corridor, target-recognition constraints, no-strike areas, and abort rules if it loses navigation confidence or enters a protected area.
At the company or battalion level, GEOINT for autonomous systems may support swarms that combine reconnaissance, strike, relay, decoy, mobility, and recovery functions. A challenge becomes coordination. Geography defines unit boundaries, maneuver corridors, target engagement areas, airspace limits, sensor handoff rules, launch and recovery sites, resupply points, communications fallback areas, and conditions for continuing, rerouting, waiting, aborting, or engaging.
At the brigade level, the problem begins to enter the realm of operational design. GEOINT for autonomous systems must support multiple autonomous formations across a larger battlespace and across multiple domains. This requires shared geospatial data, shared definitions, fratricide controls, target libraries, kill-web deconfliction, electronic warfare and position, navigation, and timing risk overlays, logistics geography, civilian-risk layers, protected-site constraints, and dynamic updates. In this scenario, the GEOINT professional is not only conducting analysis for commanders, but also encoding the geospatial logic by which autonomous forces maneuver, sense, avoid risk, and fight.
This is the heart of autonomous GEOINT—translating the commander’s intent and concept of operations into geographic knowledge that machines can execute. The commander’s tacit and explicit knowledge must be captured to train autonomous systems where to go, where not to go, what to sense, what to ignore, how to engage, what to protect, when to ask for human confirmation, and when to stop. Human command remains essential, but much of the sensing, routing, timing, prioritization, and cueing will shift to algorithms that need geography and intelligence before and during the mission.
Autonomous warfighting will change GEOINT’s products. In a machine-speed fight, the product will often not be a map, report, graphic, or briefing. It may be a training data set, machine-readable boundary, no-strike constraint, geofence, route-risk model, target-confidence score, uncertainty layer, geographic abort condition, or preloaded rule that helps an autonomous system act within the commander’s intent. Traditional products will remain, but they will be joined by products that look more like software, rules, and decision logic.
The analyst’s role will change as well. Analysts will still interpret imagery, identify features, and create intelligence, but they will also design, train, validate, monitor, and challenge AI-enabled workflows. They will increasingly supervise analytic systems and insert human judgment where it matters most. The key question will be how to divide work between people and algorithms when some analysis must be accelerated using machines and some analysis should not be delegated.
What GEOINT Must Do Now
The community of GEOINT practitioners should act now in six areas. First, it needs concepts and doctrine for autonomous GEOINT. The community must define how geographic knowledge is encoded, validated, updated, and governed for autonomous and semiautonomous systems.
Second, the GEOINT enterprise needs faster feedback loops. Field lessons, model performance, collection gaps, jamming effects, and operator observations must quickly inform mission data, methods, doctrine, training, education, and procurement.
Third, GEOINT professionals must become more comfortable with low earth orbit analysis, nonliteral sources, and sensor geography, including the electromagnetic landscape. Relevant spatial evidence will increasingly appear in emissions, anomalies, relay links, navigation failures, and adaptation patterns.
Fourth, the GEOINT community must treat requirements, production, and procurement as intelligence problems. The side that can envision, decide, learn, produce, train, and adapt faster will gain the advantage. Readiness is therefore an intelligence problem as well as a logistics and training problem. This requires fixing a basic weakness in the requirements process. Government agencies often wait for military services to issue new requirements before changing products. That approach is no longer sufficient. GEOINT practitioners must anticipate future needs and reimagine products and services now. The community needs ways to accelerate bottom-up innovation, reduce acquisition barriers, and bypass procurement processes too slow for battlefield change.
Fifth, the GEOINT aperture must widen. Orbital systems in low earth orbit, undersea and subterranean domains, new chokepoints, electromagnetic terrain, cyber infrastructure, and cognitive warfare all have geographic dimensions of strategic and operational importance.
Sixth, GEOINT training, education, and learning culture must change. The community needs leaders and analysts who understand geography, autonomy, data, AI, electronic warfare, position, navigation, and timing, and how to encode the commander’s tacit and explicit geographic knowledge. They must know how to work with cognitive physiologists, software engineers, and data scientists and when to slow the machine down because the judgment is too consequential to delegate.
The autonomous battlefield is here. This moment matters. Recent conflicts show that warfare is becoming faster, more transparent, more autonomous, more distributed, and more adaptive. The concern is whether GEOINT organizations can keep pace while maintaining human judgment, skill, accountability, and legitimacy.
The warfighter cannot afford for the GEOINT community to be horse holders. Successful autonomous systems need the commander’s explicit and tacit knowledge of the mission’s intent and concept of operations translated into geographic knowledge that machines can execute. This requires the GEOINT community to accept the inevitability of autonomous systems and rethink its current enterprise supporting the warfighter.
Developing this technology matters, but the harder problem will likely be human. People, doctrine, organizations, education, training, and leadership will determine whether autonomous technology is a decisive advantage.
Todd S. Bacastow is a Penn State academy professor and teaching professor emeritus in Penn State University’s College of Earth and Mineral Sciences, where his focus has been on emerging technologies and the education of geospatial analysts for the US intelligence community. Before joining Penn State in 1994, he served in the US Army in infantry and engineer assignments, taught at the US Military Academy, and retired from military service. He is a 1974 graduate of the US Military Academy and holds an MS and PhD from Penn State.
Author’s note: While I accept responsibility for any flaws in this manuscript, I could not have produced it alone. The ideas grew out of many conversations with colleagues and my son. I especially thank the “Old P,” Gilbert W. Kirby, Jr., for his mentorship early in my career, and my POTC ’74 classmate, David Petraeus, whose article helped me develop a clearer vision of GEOINT’s future.
The views expressed are those of the author and do not reflect the official position of the United States Military Academy, Department of the Army, or Department of Defense.
Image credit: Sgt. Alfonso Livrieri, US Marine Corps
