- DK NEJET
The Illusion of a Magic Button: Why Autonomy Alone Won’t Change the Battlefield

The European defense market is pouring money into autonomous systems, which is good. However, the danger lies in a seductive myth that all it takes is to buy the technology, hand the military a “box with a magic button,” and watch it solve manpower shortages, electronic warfare, and the need for faster decisions.
The truth is less glamorous – the autonomy itself, for all its benefits, is not a solve-it-all solution because it is only as effective as the people, processes, and systems around it.
So imagine we handed you a box with an “AUTONOMY” button. Then what?
- Who defines the mission?
- Who sets the constraints?
- Who checks the system’s decisions?
- Who steps in when the situation changes?
- And who carries the responsibility for the outcome?
- Analyzes large data volumes
- Responds in real time
- Operates without fatigue
- Coordinates multiple platforms
- Apply contextual judgment
- Recognize ambiguity
- Adjust priorities on the fly
- Decide with incomplete data
- whether it’s a friendly aircraft’s flight that was planned;
- whether another group’s route has changed;
- whether the recognition system’s data is up to date;
- whether the object has been captured or reprogrammed;
- whether the line of engagement has changed since the mission was assigned, and so on.
- how many people are currently in the area where debris might fall;
- whether a concert or soccer game is taking place along the trajectory;
- whether it is possible to track the target and intercept it later over safer territory;
- which poses a greater risk—allowing the object to continue its flight or destroying it immediately;
- what the secondary consequences of an interception would be.
- what the system does after losing communication;
- how long it continues the mission;
- under what conditions it returns;
- what actions it is prohibited from performing without confirmation;
- what happens if the situation differs from the original plan.
- how roles are distributed among the platforms;
- which one is responsible for surveillance;
- which one acts as a relay;
- how the system reroutes in case one of the drones is lost;
- how to avoid overlapping targets;
- what to do after a mission is canceled;
- at what point the operator must intervene.
- operator training;
- training for commanders;
- practical drills for non-standard scenarios;
- clear rules and tactics for use;
- regular software updates;
- a channel for rapid feedback between the user and the developer.
- Technology—platforms, sensors, software, and algorithms.
- Trained personnel—operators, commanders, instructors, and technical teams.
- Doctrine of use—clear rules, scenarios, and limits for autonomous decisions.
- Integration—linkage with intelligence, command systems, and other platforms.
- Continuous feedback—the ability to quickly translate field experience into system updates.
- Performance metrics—not the number of features in a presentation, but mission completion time, operator workload, resilience to electronic warfare, and actual effectiveness.
- What happens when the system is unsure?
- At what point does it contact the operator?
- What happens after a loss of communication?
- How can the operator modify or cancel a mission?
- How does the system distinguish a friendly target from a hostile one?
- Who trains people to use it, and how?
- How does combat experience feed back into the development team?
Autonomy is NOT the absence of humans
The main misconception is to view autonomy as the complete removal of humans from the process. However, in real combat conditions, even the most sophisticated system operates with incomplete, contradictory, or outdated information. We’re stating a fact here, but well, such is real life.
That is precisely why the future lies not in a “human out of the loop” model, but in properly structured human-machine interaction.
Division of strengths
Where AI and humans each perform best
Two different kinds of capability — sharpest when they overlap.
Machine speed under human judgment — accountable decisions, even when the picture is incomplete.
AI excels at analyzing large data volumes, responding rapidly, operating without fatigue, and coordinating multiple platforms. Humans excel at contextual judgment, recognizing ambiguity, adjusting priorities, and making decisions when data is incomplete.
Popular culture has explored this tension. In Person of Interest, The Machine identifies patterns and flags possible threats, but human characters still interpret the context and decide what to do. By contrast, 2001: A Space Odyssey’s HAL 9000 shows what can happen when an AI must reconcile conflicting instructions without meaningful human oversight. The lesson is not that AI must be passive, but that machine analysis and human judgment work best together.
Autonomy should transform operators’ roles from manual control to objective-setting, constraint monitoring, and critical intervention.
In this article, we’ll take a look at five examples from our operators’ experience in Ukraine – what an operator should take into consideration when flying over contested environments and why a “magic pill” won’t work.
Example 1. Gray zones
Gray zones are areas where both enemy and friendly systems might be operating at the same time. A drone detects an object that resembles a target in terms of shape, speed, or thermal signature.
The algorithm may correctly classify the type of object, but may not know the full operational context:
The list of these questions can be endless.
In such a situation, the autonomous system should not try to “guess,” but rather determine the level of uncertainty and pass the decision on to the operator. Sometimes the best decision for the system is not to act on its own.
Example 2. Unknown object over civilian territory
Another situation happens when, for example, there’s an enemy drone that has to be downed because its goal is critical infrastructure, but there are civilian objects below.
An autonomous system detects an unknown aerial object whose trajectory and behavior match the profile of a potential threat. Based on its technical parameters, the system can classify it as dangerous and even calculate the optimal interception point.
However, a technically feasible interception does not always mean it is the right decision.
The object may be flying over a stadium, a transportation hub, a residential neighborhood, or the site of a mass gathering. Even a successful strike in such an area could cause debris to fall, fuel to explode, or the payload to detonate. As a result, an action that formally neutralizes the threat could lead to significantly more civilian casualties.
The algorithm detects the object’s characteristics, its speed, direction of travel, and the likelihood of a threat. However, it may not have the full context:
Therefore, in the context of civilian infrastructure, the autonomous system should not decide on its own whether to destroy the target, but rather assist the operator in quickly assessing the scenarios: determining the threat level, predicting the trajectory, calculating the potential debris impact zone, and suggesting the safest moment for intervention.
The final decision must be made by a trained operator who sees the broader operational and humanitarian context because speed of interception isn’t always the best criterion. That’s one of the reasons we build our platforms with a disarming option – in case the situation changes, the armed drone can be disarmed and it can return to its starting point.
This example illustrates a key limitation of autonomy: the system can execute a command flawlessly, but only a human can assess whether a successful interception would cause an even greater tragedy than the threat itself.
Example 3. Loss of communication doesn’t mean the mission’s aborted
The platform received its mission assignment, but while carrying it out, it came under the influence of electronic warfare (EW) measures and lost communication with the operator.
A simple remotely controlled system becomes uncontrolled. An autonomous system can continue along its route using inertial navigation, optical flow, or visual odometry. It can also return, enter standby mode, or carry out a predefined portion of the mission.
But this only works if the team has defined the rules in advance:
Autonomy does not eliminate the need for planning. On the contrary, it requires significantly higher-quality planning.
Example 4. One Operator and a Swarm
In a presentation, the ability of a single operator to control dozens of drones looks like technological magic, and often people expect the swarm to work on its own. However, simply increasing the number of platforms does not create a combat advantage.
If the interface forces a person to manually control each platform, it is not a swarm, but rather dozens of individual drones on a single screen. The operator quickly becomes the “bottleneck” of the entire system.
Within our swarm system, the majority of drones remain in a ready state, hovering in the air. Should the designated strike drone be compromised, the system reconfigures automatically, minimizing operator mental load by reducing the number of drones that require active monitoring.
Swarm architecture
Hold, strike, reconfigure
Most drones wait in readiness while one engages. If the lead is lost, the swarm re-tasks itself — so the operator only watches a few.
Readiness. Most drones hold position, primed to act.
Strike. A single primary drone engages the threat.
Reconfigure. If the primary is lost, a ready drone takes over automatically.
Low load. The operator only monitors the lead drones.
To have a true swarm, it is necessary to determine:
True autonomy lies not in the number of drones, but in how many decisions the system can safely make on its own—and how clearly the operator can see what requires their attention.
Example 5. The Right Technology Without the Right Training
A European unit receives a state-of-the-art autonomous platform which is capable of operating without GPS; it can recognize objects and adjust its route. However, the military personnel have trained primarily on systems that require constant manual control.
As a result, they either do not use the autonomous functions at all or apply them in scenarios for which they were not designed. In both cases, the technology’s actual effectiveness will be significantly lower than advertised.
Delivering a system without training is like delivering an airplane without crew training. Having the technology does not automatically mean having the capability.
What is needed is not just product documentation, but:
At DK NEJET, we continually talk about the fact that we engineer (R&D and software), we build (the physical platforms), and we train operators that can use our hardware and software to its full potential.
Autonomy is a system, not a feature
It’s not possible to simply add autonomy to an existing platform and expect a systemic advantage. It must be integrated into the entire operational cycle, and if any part of this cycle fails, autonomy loses a significant portion of its value.
For example, a system may be excellent at detecting targets but, without up-to-date data, may set priorities incorrectly. It may follow a route precisely but not be integrated with the command system. Or it may demonstrate high performance on the test range but lack a mechanism for rapid updates once new countermeasures emerge.
Europe needs capability, not just a “box with a magic button.”
The Ukrainian experience has shown that technology cannot be a finished product in the traditional sense. The battlefield changes faster than the traditional procurement cycle.
What worked yesterday may lose its effectiveness tomorrow due to new electronic warfare frequencies, changes in tactics, the emergence of new target profiles, or new methods of camouflage. Therefore, true autonomous capability includes:
The Key Question for Buyers
When choosing an autonomous system, it’s not enough to ask, “What can it do on its own?”
Much more important questions are:
Conclusion
There’s no magic button. (Sorry, if you were expecting one.)
Autonomy will indeed change warfare, but not because it will allow us to eliminate the human element. It will change warfare by enabling a single, well-trained person to control a more complex system, make decisions faster, manage more platforms, and put their life at less risk.
Here’s to a safer future for us all.

