SYNOPSIS
This article explores the necessity and opportunities of pervasive Human-Machine Teaming (HMT). It delves into the requirements for efficient HMT broadly deployed across all echelons and examines the technical challenges, primarily focusing on telecommunications and associated video networking challenges.
ACRONYMS
- MANET: Mobile Ad Hoc Network
- MOSA: Modular Open Systems Approach (DoD)
- NGVA: NATO Generic Vehicle Architecture
- QoS: Quality of Services
- SATCOM: Satellite Communications
- STANAG: STANdardization AGreement (NATO)
- VICTORY: Vehicular Integration for C4ISR/EW Interoperability (DoD)
Introduction
In modern warfare, battlefields generate overwhelming volumes of data, with estimates indicating the scale of sensor data alone can exceed petabytes daily. This data overload necessitates AI-enabled systems to filter and prioritize actionable intelligence efficiently, improving operational decision-making. AI applications have demonstrated remarkable capacity to reduce decision-making time by significant margins, sometimes by up to 70%, thereby enhancing situational awareness and operational efficiency in high-tempo environments. Such capabilities provide commanders with the critical advantage of making faster, more informed decisions, which are decisive in contemporary conflict scenarios.
The recent operation “SpiderWeb” represents an inflection point in modern warfare — a breakthrough in tactical UAV operations with truly strategic implications. It stands as a shining example of Human-Machine Teaming (HMT) in action, where AI models, unmanned drones, and distributed human stakeholders collaborate seamlessly across distances. While HMT encompasses a broad range of use cases, its operationalization is highly dynamic, shaped by the maturity of technologies, mission-specific constraints, and evolving frameworks for responsibility between humans and machines.
This dynamic interplay is often captured through the three-tier model of HMT: Human-in-the-Loop (HITL), Human-on-the-Loop (HOnTL), and Human-off-the-Loop (HOffTL):
- Human-in-the-Loop (HITL): In high-stakes scenarios requiring precision and accountability, human operators remain directly involved in decision-making. For example, a drone operator may control reconnaissance flights while AI provides recommendations on optimal flight paths or potential targets.
- Human-on-the-Loop (HOnTL): In situations where operational tempo demands rapid decisions, humans oversee and intervene only when necessary. AI systems handle routine tasks such as monitoring multiple video feeds or coordinating swarm activity, allowing operators to focus on higher-order decisions.
- Human-off-the-Loop (HOffTL): For highly autonomous missions, such as long-range surveillance or logistics resupply, AI systems operate independently while adhering to predefined rules of engagement and mission parameters.
Pervasive Human-Machine Teaming is essential for achieving decision dominance in future conflicts. To truly realize the potential of such tactical systems, their deployment must be pervasive — integrated at scale across all levels and domains of the battlespace. However, enabling this shift comes with a set of formidable technical challenges. Chief among them is the connectivity infrastructure, which must support resilient, high-performance HMT under the fragmented and contested network conditions typical of modern combat environments. This paper focuses on addressing that foundational challenge.
(1) The necessity of pervasive tactical HMT
Modern military operations are increasingly shaped by the principle of decision dominance: the ability to consistently make faster, better, and more resilient decisions than the adversary. Achieving this requires applying the concept of subsidiarity — empowering decisions to be made at the lowest possible echelon, closest to the tactical edge, while ensuring coherence with higher-level intent.
In today’s environment, decision-making is inseparable from interacting and collaborating with unmanned systems — whether to collect intelligence, extend reach, or execute effects. This is the essence of Human-Machine Teaming (HMT). The challenge is not confined to a single echelon; It spans a wide spectrum of use cases and mission scenarios, encompassing multiple forms of Human-Machine Teaming — from autonomous sensing, to decision-support functions, through to direct action — and aligning with the three commonly recognized tiers (HITL, HOnTL, HOffTL).
One of the clearest shifts since earlier doctrinal writings on HMT is the move away from a single, linear OODA loop in which AI merely supports the human at the Observe or Orient phases. On today’s battlefield, humans and machines operate through multiple concurrent OODA loops distributed across sensors, shooters, and decision nodes. In this environment, the human role is less about single-point control and more akin to that of a conductor orchestrating an orchestra of machine-augmented micro-loops. Rather than personally deciding and acting on every tactical event — a task that would quickly exceed human cognitive bandwidth — the commander aligns the collective behaviour of AI systems towards an overarching operational end-state. This orchestration model preserves the human as the author of intent and the arbiter of acceptable risk, while allowing AI systems at the edge to execute bounded Decide/Act functions at the speed required to survive in hyper-tempo engagements.
To enable every echelon to act decisively, there is a necessity to establish pervasive HMT. Without it, the full potential of unmanned systems remains fragmented. With it, commanders and squads can harness information, synchronize actions, and achieve outcomes that decisively shift tactical advantage.
Pervasive HMT is not static. It unfolds in a dynamic operational and technological environment: doctrines are evolving, unmanned systems are rapidly upgraded, AI models are continuously trained and improved, and battlefields are shaped by bottom-up innovation and fast iterative engineering. In such a fluid setting, any closed or rigid solution risks near-immediate obsolescence. What is required instead are open, adaptable, and scalable architectures — systems that can absorb a huge amount of visual information, new sensors, integrate new algorithms, and evolve in step with operational innovation.
A compelling illustration of this necessity is the concept of an all-in-one HMT solution for deployed squads. Within minutes of insertion, a squad could launch drones, operate them from a central interface, apply AI models to extract actionable insights, and collaborate around a shared tactical picture that fuses all sensor inputs. This capability enables rapid situational understanding, faster decisions, and synchronized actions that multiply combat effectiveness.
In short: efficient, pervasive HMT at the tactical edge is vital to fully exploit the advantages of unmanned systems and deliver decisive impact in future operations. It ensures subsidiarity by empowering squads to act, and it delivers decision dominance by compressing the time between sensing, deciding, and acting.
(2) The key components of Tactical HMT
Pervasive tactical Human-Machine Teaming (HMT) achieves operational effectiveness by seamlessly integrating critical components that enable synchronized collaboration, real-time decision-making, and enhanced situational awareness. These components must work in unison to address the challenges posed by modern combat environments, where the volume of data, speed of operations, and complexity of scenarios increasingly exceed unaided human capabilities. Importantly, tactical HMT must support varying degrees of human involvement — whether human-in-the-loop (HITL), human-on-the-loop (HOnTL), or human-off-the-loop (HOffTL) — to ensure adaptability to mission-specific requirements. This section explores these foundational elements, with a focus on overcoming operational challenges through advanced technologies and AI-driven solutions.
The Pivotal Role of Visual Information
In the tactical domain, visual information is the backbone of decision-making and operational awareness. A single image, video feed, or tactical overlay can deliver actionable intelligence far more efficiently than textual reports or verbal communication. However, as the deployment of drones, sensors, and unmanned systems expands, the sheer volume of raw visual data generated across the battlespace can quickly overwhelm human operators. In scenarios involving multiple drones or even swarms, the difficulty of manually analyzing and correlating live feeds in real time becomes a critical bottleneck.
AI-driven systems offer a transformative solution by automating the extraction and transformation of relevant information from raw visual data. For example, AI can identify and classify objects of interest, track targets across multiple feeds, and even correlate visual inputs with metadata, such as timestamps or geospatial coordinates. These capabilities not only reduce the cognitive burden on operators but also ensure that decision-makers can focus on actionable insights rather than being inundated by irrelevant or redundant data.
From Cloud to Edge: Managing the Data Deluge
The enormous volume of data generated by tactical HMT systems must be transported, processed, and delivered in real time to support operational needs. While cloud-based processing offers significant computational power, it relies on dependable network connectivity — a resource that is often scarce or contested in combat environments. Tactical networks face challenges such as bandwidth constraints, latency, and intermittent availability, all of which impact the timely processing and delivery of critical information.
Edge deployment provides a practical remedy to these challenges by enabling data processing closer to the point of origin. AI models deployed at the edge can filter, prioritize, and analyze data locally. This reduces the need to transmit raw data over bandwidth-limited networks. For instance, rather than streaming all video feeds to a central server, edge AI can identify and transmit only the most relevant segments, such as footage showing potential threats or mission-critical events. This approach not only alleviates network congestion but also ensures that operators receive actionable insights faster, even in degraded communication environments.
AI for Situational Analysis & Data Transformation
Beyond filtering and prioritizing data, AI plays a crucial role in extracting and transforming information into actionable formats. For example, AI can transcribe voice communications into text, enabling rapid dissemination of critical updates across distributed teams. Similarly, AI-driven situational analysis can integrate data from multiple sources — such as live video feeds, geospatial overlays, and sensor inputs — to generate a coherent operational picture.
These capabilities are particularly valuable in complex, dynamic environments where human operators may struggle to process and correlate information in real time. By automating these tasks, AI not only enhances situational awareness but also ensures that critical insights are delivered to the right stakeholders at the right time.
Trust, Communication &Training
Trust remains central to whether pervasive HMT can succeed. Large language models and dialogue-based AI systems alter the psychology of teaming: their human-like responses build an immediate sense of rapport and reliability. Yet this trust is fragile — operators forgive repeated human mistakes more easily than a single visible machine error. As a result, training must shift to include interaction techniques as a soldiering skill.
Prompt engineering, far from being a programming trick, becomes a practical method of communicating intent, constraints, and urgency to decision-support systems. Soldiers who can articulate precise requests, such as framing context, specifying output formats, and signalling confidence thresholds, will gain more accurate and actionable assistance. Simulation is crucial here, not only for training humans in these new habits of communication, but also for producing the synthetic data needed to refine AI models in environments where real-world data is scarce or ethically difficult to obtain.
Unified Visual Ecosystem & Information Fusion
A unified visual ecosystem is critical for ensuring that all stakeholders within the tactical HMT framework operate from a common operational picture. This ecosystem must integrate data from diverse sources, including drones, sensors, and ground units, into a single, coherent interface. AI-driven information fusion plays a key role in this process, correlating disparate data streams to provide a comprehensive view of the battlespace.
For instance, live drone feeds can be overlaid with geospatial data, tactical maps, and metadata to create an enriched visual interface. Operators can interact with this interface to explore specific areas of interest, review historical data, or monitor real-time events. By presenting relevant information in a unified format, tactical HMT systems enhance situational awareness and enable more effective decision-making at all levels of command.
Reducing Cognitive Overload Through AI
Finally, one of the most critical contributions of AI in tactical HMT is its ability to reduce cognitive overload on human operators. By automating routine tasks, filtering irrelevant data, and providing actionable insights, AI enables operators to focus on the most critical aspects of their mission. This not only improves decision-making but also enhances operator effectiveness and endurance in high-stress environments.
Real-Time Decision-Making & Collaboration
The pace of modern warfare demands rapid and coordinated decision-making, often in environments where seconds can determine mission success or failure. Tactical HMT systems must support real-time collaboration between humans and machines, enabling distributed teams to assess, decide, and act with unprecedented speed. Features such as synchronized video playback, shared tactical overlays, and collaborative annotation tools enhance coordination and ensure that all stakeholders remain aligned on mission objectives.
For example, a forward-deployed platoon leader viewing a live video feed from a drone can annotate potential threats directly on the feed, with those annotations instantly visible to a remote commander. This level of real-time collaboration ensures that decisions are based on a shared understanding of the operational picture, reducing the risk of miscommunication or delays.
Bounded Autonomy & Accountability
A pervasive deployment of HMT requires robust frameworks for bounding machine action without suffocating its utility. Here, the analogy to rules of engagement (ROE) is powerful: just as soldiers are authorised to act within clearly specified parameters, so too can AI be constrained by geofences, target classes, and permissible effects. Within these envelopes, machines may interpret and act with high speed and unpredictability, providing the strategic edge of surprise against adversary systems.
Accountability follows the same logic as with human subordinates. Responsibility remains with the commander who delegates lethal authority, provided due diligence has been exercised in testing, training, and assigning the system. The critical challenge is not inventing a new legal architecture, but embedding AI systems within the well-established traditions of military command responsibility while ensuring audit trails and post-mission provenance to support review and lessons learned.
Resilient Communications for Tactical Networks
The effectiveness of tactical HMT depends on the resilience of its communication infrastructure. Real-time visual collaboration and synchronized decision-making require a robust network that can withstand the challenges of contested and bandwidth-constrained environments. Tactical HMT systems must incorporate advanced communication technologies, such as adaptive video codecs, automated Quality of Service (QoS) management, and multi-path networking, to ensure reliable performance under adverse conditions.
In conclusion, the key components of tactical HMT — ranging from visual information dominance to edge AI deployment — form the foundation for achieving decision dominance in modern combat environments. By integrating these elements into a cohesive framework, tactical HMT systems can overcome the challenges of data overload, bandwidth constraints, and operational complexity, ensuring that human operators and machine systems work together seamlessly to achieve mission objectives.
(3) The communication Infrastructure challenges
As described above, Pervasive HMT requires supporting large-scale, real-time, synchronized, and interactive visual communications over tactical networks. In combat environments where scarcity of power, compute, and bandwidth is the norm, these sophisticated visual communications present a number of critical challenges that must — and can — be addressed.
While the Ukrainian battlefield provides valuable insights, its communications model is mostly cloud-based and LEO-centric — heavily reliant on Low Earth Orbit satellite networks like Starlink — and may not fully reflect the operational constraints of other European armies, which depend on a heterogeneous mix of network technologies (e.g., HF, SATCOM, terrestrial mobile) to achieve redundancy and wider geographic coverage across coalitions.
Bandwidth Availability & Fluctuations
Automated bandwidth and Quality of Service (QoS) management mechanisms must be more broadly adopted at the transport layer (OSI Layer 4) to ensure resilient, real-time visual communication across heterogeneous networks. These mechanisms already exist and can be deployed over existing infrastructure, enabling dynamic video codec adaptation to fluctuating network conditions — via resolution, frame rate, or bitrate adjustment — and delivering high visual quality with robust resilience, especially in contested or bandwidth-constrained tactical environments.
Latency & Synchronization
Most current streaming solutions rely on buffering to compensate for jitter and packet loss, but this approach introduces latency that is detrimental to real-time collaboration and synchronized AI workflows. Ultra-low-latency video systems are essential to support applications such as remote teaming, AI inference-on-video, and autonomous system coordination.
New video technologies — particularly GPU-based transcoding — now enable complex processing with very low latency while also supporting precise stream synchronization. This is especially critical for data augmentation use cases, where accuracy is significantly improved by the simultaneous analysis of multiple synchronized video streams.
Scalable & Redundant Architectures
Resilient mesh networks, supported by MANET technologies and Layer 3 routing, ensure reliable connections. These networks can be enhanced with software-based tools that manage media continuity, load balancing, and automatic failover. These technologies are entirely software-based, demand low computational power, and can be embedded in current generations of telecom devices or edge servers.
Jamming & Security
Operational environments may also experience deliberate electromagnetic interference (jamming) and other physical-layer attacks that can deny or severely degrade individual links. HMT systems should therefore be designed on the assumption that any single bearer can become unavailable at short notice — and that application continuity must nevertheless be preserved. Practically, this means relying on a combination of communication technologies (HF, SATCOM, terrestrial cellular, MANET/mesh) and placing a lightweight middleware layer between network transports and HMT applications.
That middleware transparently manages session continuity, authentication, and encryption while performing rapid path selection and failover (e.g., switching from SATCOM to a terrestrial link), adaptive flow control, and codec adaptation so that upper-layer apps keep working without embedded reconfiguration. By abstracting connectivity in this way, operators preserve real-time visual collaboration and AI workflows even under contested spectrum conditions, while enabling policies that prioritize and gracefully degrade non-critical video streams.
Forward Compatibility & Simplified Integrations
The diversity of video systems deployed in theater has led to significant codec interoperability challenges. Different video codecs are optimized for varying latency, bandwidth, and power constraints, making seamless integration complex.
To address this, high-performance, real-time transcoding should be deployed at the network edge to ensure compatibility and forward interoperability between systems and sensors. This approach enables systems to remain flexible and rapidly adaptable to constantly evolving integration needs. For example, platoons often need to integrate new drones using different video codecs. Supporting them with codec-agnostic solutions enables fast, “plug-and-go” integration — critical for maintaining operational tempo. Finally, solutions must interoperate using STANAGs, NGVA, VICTORY, or MOSA principles.
Edge-Based Filtering, Prioritization &Overflow Mitigation
The volume of visual data generated across the battlespace is staggering and rapidly increasing. Even with continued investment in infrastructure, network saturation is inevitable without intelligent data management. AI-based video filtering, prioritization, and routing at the edge will be essential to avoid overwhelming networks and operators.
These capabilities — often referred to as Visual AI Routing — are already being discussed as part of 6G network specifications for defense and public safety use cases. 6G is a long shot, but solutions exist today to implement edge AI filtering and routing using lightweight or intermediate models at the edge. This is another key benefit of implementing real-time video transcoding at the edge or in the network.
Energy & Compute Power Scarcity
As edge AI becomes more prevalent, managing power consumption at the edge is increasingly critical. While new generations of GPUs offer improved performance-per-watt, additional strategies are necessary to balance energy use and computational demand.
A practical approach involves selecting the most appropriate dataset and inference model for each use case. Tasks requiring lower precision can operate on encoded visual streams using lightweight, compute-efficient models — preserving energy for more demanding operations that require decoded video streams and high-accuracy inference.
Another key consideration is the energy footprint of tablets and mobile devices used by soldiers. To extend battery life, it is essential to reduce the number of video streams these devices must decode and compose. A highly effective strategy is to send a single composed video stream that aggregates multiple sources (e.g., video feeds, dynamic maps, metadata) into one unified view, minimizing decoding load while maintaining situational awareness.
Security
Security must be embedded across all layers of the visual communication infrastructure. In contested environments, video and control data streams are vulnerable to interception, spoofing, and disruption. Encryption, stream authentication, and secure key exchange must be enforced across edge and network components — without introducing additional latency. Furthermore, integration of zero-trust principles and dynamic threat detection at the edge will be key to maintaining confidentiality, integrity, and availability in highly distributed and mobile C2 systems.
Open Architecture
Finally, an open architecture is fundamental for building resilient, scalable, and adaptive systems. At the heart of such architectures is the network layer, enabling seamless integration of an unbounded ecosystem of sensors, devices, applications, and AI models. As AI evolves rapidly, operators must be able to dynamically link models to any live or recorded visual stream in the network. The ability to connect specific AI inference engines directly to live video streams — without requiring deep integration — is becoming a foundational capability for modern C2 and ISR systems. Multiple use cases are already discussed associating multiple and staged AI models.
Conclusion
Pervasive tactical Human-Machine Teaming (HMT) is no longer a distant aspiration — it is a necessity for modern forces seeking decision dominance in highly dynamic, contested environments. Ignoring this opportunity risks operational inefficiency or even decision paralysis. The operational and technological trends are clear: future battles will be defined by the speed and quality of distributed decisions, enabled by seamless human-machine collaboration across every echelon.
To achieve this, defense organizations must embrace architectures that are open, adaptive, and resilient. The challenges are significant — bandwidth scarcity, latency, jamming, interoperability, and energy constraints — but none are insurmountable. Advances in edge AI, real-time transcoding, adaptive networking, and visual information fusion already provide the building blocks for resilient, scalable, and secure HMT deployments.
What is required now is a mindset shift: from siloed and rigid solutions to a pervasive, integrated ecosystem where unmanned systems, AI models, and human operators collaborate in real time, regardless of network conditions. Such an ecosystem will not only unlock the latent potential of today’s technologies but also ensure forward compatibility with tomorrow’s innovations.
Ultimately, the ability to deploy pervasive HMT at scale will determine which forces can consistently out-think, out-decide, and out-maneuver their adversaries. By confronting the network technology challenges head-on and investing in open, flexible, and future-proof systems, modern militaries can ensure that human-machine teams become a decisive instrument of strategic advantage.
As Human-Machine Teaming continues to evolve, the seamless collaboration between humans and machines will enable faster, more accurate decision-making in complex, high-pressure environments, fundamentally redefining how teams operate under uncertainty. By embracing open architectures, edge AI, and adaptable systems today, organizations can position themselves at the forefront of this transformation, ensuring a competitive advantage not just in defense, but in every domain where timely, informed decisions are paramount. The future of HMT is not limited to tactical superiority — it is a blueprint for a new era of human-machine synergy across society.
Executive Takeaway
Pervasive tactical Human-Machine Teaming (HMT) is essential for achieving decision dominance in future conflicts. Success will depend on resilient, open, and adaptive network architectures capable of sustaining real-time, visual collaboration under contested conditions. By integrating edge AI, adaptive networking, and scalable visual ecosystems, forces can ensure that human and machine teams operate seamlessly, remain future-proof, and deliver decisive operational advantage.
References
Pelicano, M. And Duff, D. (2018). Sensor overload. The United States Army. https://www.army.mil/article/199871/sensor_overload (accessed October 3rd, 2025)
Collins, S. (2025). Army senior leader outlines data-driven transformation for warfighting superiority. Army.mil. https://www.army.mil/article/288157/army_senior_leader_outlines_data_driven_transformation_for_warfighting_superiority (accessed October 3rd, 2025)
Baxter, C.A. (2024). Enhancing tactical level targeting with artificial intelligence. Line of Departure Journal. https://www.lineofdeparture.army.mil/Journals/Field-Artillery/FA-2024-Issue-1/Enhancing-Tactical-Level-Targeting/ (accessed October 3rd, 2025)
Bondar, K. (2024). How Ukraine’s Operation “Spider’s Web” Redefines Asymmetric Warfare. Center for Strategic and International Studies. https://www.csis.org/analysis/how-ukraines-spider-web-operation-redefines-asymmetric-warfare (accessed October 3rd, 2025)
Bondar, K. (2024, July 8). How Ukraine’s Operation “Spider’s Web” Redefines Asymmetric Warfare. Center for Strategic and International Studies. https://www.csis.org/analysis/how-ukraines-spider-web-operation-redefines-asymmetric-warfare (accessed October 3rd, 2025)
CONTRIBUTED BY Dorottya Zsiborács
Dorottya is a defense tech specialist with experience at Karve International, a strategic national security growth consultancy, and Helsing, a leading defence AI company in Europe. She has led growth strategies for AI-enabled systems across UK, NATO, and AUKUS markets, with a focus on operational deployment and mission alignment. An Oxford MSc graduate, who researched human-machine teaming in warfare, she brings technical, industry and policy experience.
CONTRIBUTED BY Kurt J. Zuckermann
Kurt has extensive experience working for technology companies in the fields of ICT and AI, specializing in mission-critical applications including air traffic management, public safety, and defense. At eyeson, he focuses on driving innovation through cutting-edge applications and collaborative R&D initiatives. Recognizing the transformative potential of integrating AI with *eyeson One View* to enhance situational awareness and decision superiority, he is dedicated to developing projects that bring this vision to life.
