.svg){kind=icon}
Learn
The kill chain is the military targeting sequence - classically find, fix, track, target, engage, and assess (F2T2EA) - that takes a force from detecting a potential target to acting on it and judging the result.
Last updated: October 23, 2026
Why it matters to your mission. A target is only useful if you can act on it before it moves or the moment passes. The kill chain is the sequence that gets a force from "something is out there" to "we engaged it and here is the result." Every stage before the act of engaging is, underneath, a data and correlation problem: detecting, locating, identifying, and tracking a target all depend on fusing fragmented sensor and intelligence data into one trusted picture. The chain is only as fast as the data feeding it, which is why closing it at machine speed is a reasoning-infrastructure problem, not a weapons problem.
The kill chain is the military targeting sequence - classically find, fix, track, target, engage, and assess (F2T2EA) - that takes a force from detecting a potential target to acting on it and judging the result. It is the condensed, execution-level process a force runs to prosecute a target that was not fully planned in advance, and the steps can run in sequence or overlap as the situation changes.
In U.S. joint doctrine the construct is formalized as the dynamic targeting process. Joint Publication 3-60, Joint Targeting describes the fifth phase of the joint targeting cycle as a short run of steps and states plainly that "dynamic targeting has often been called 'F2T2EA' (find, fix, track, target, engage, and assess) or 'the kill chain.'" The same publication is explicit that the steps "may be accomplished iteratively and in parallel," so the popular image of a strict linear chain is a simplification of how the process actually runs.
Key takeaways
F2T2EA has six stages. The first five - find, fix, track, target, engage - are what JP 3-60 places inside phase 5 of the joint targeting cycle; the sixth, assess, is the cycle's final phase, "common to both deliberate and dynamic targeting." The doctrine's one-line descriptions make the pattern clear: four of the six stages are about seeing and understanding the target, and only two are about deciding and acting.
| Stage | What happens (per JP 3-60) | The data problem underneath it |
|---|---|---|
| Find | "Possible targets are detected and classified for further prosecution." | Pulling a real detection out of noisy, multi-source sensor and intelligence feeds, and classifying it correctly. |
| Fix | "Actions to determine the location (fix) of the potential target." | Resolving a precise, trusted location by correlating reports that disagree, lag, or use different reference frames. |
| Track | "The target is observed and its activity and movement are monitored." | Maintaining one continuous track as the target moves and as different sensors cue in and drop out. |
| Target | "The decision is made to engage the target... and the means to do so are selected and coordinated." | Fusing the track with identity, context, rules, and risk so the decision is defensible, not just fast. |
| Engage | "Action is taken against the target." | Getting a complete, current picture to the acting element and confirming it is still valid at release. |
| Assess | The assessment phase "examines the results of the target engagement." | Correlating post-action reporting across sources to judge effect and decide whether to re-engage. |
JP 3-60 notes the division of labor directly: "the find, fix, and track steps tend to be ISR-intensive, while the target and engage steps are typically labor-, force-, and decision-making intensive." In plain terms, most of the chain is an intelligence, surveillance, and reconnaissance problem. The chain does not begin with a weapon; it begins with a question about data - is there a target, where exactly is it, and can we trust the track enough to act.
The kill chain is a concrete targeting workflow; the OODA loop is a general decision-making model. The OODA loop - observe, orient, decide, act - was developed by U.S. Air Force Colonel John Boyd as a way to describe how any competitor out-tempos an adversary by cycling through those four phases faster and with better orientation. It is cognitive and universal. The kill chain, by contrast, is a specific sequence of targeting actions - F2T2EA - for finding and engaging a particular target.
The two are complementary rather than competing. The kill chain can be read as OODA applied to the targeting problem: find and fix map to observe and orient, target maps to decide, and engage maps to act, with assess closing the feedback loop. The practical point is the same in both frames. Boyd put orientation - sense-making - at the center of the loop, and in the kill chain the sense-making stages (find, fix, track) are where most of the time and most of the data work actually live. Faster orientation, built on better-fused data, is what compresses either cycle. This is the mechanism behind decision advantage: the side that closes its loop first, with a picture it can trust, gets to act while the other side is still orienting.
Dynamic targeting is the prosecution of targets that emerge too late to be planned in advance, and it is the doctrinal home of the kill chain. JP 3-60 splits joint targeting into two categories. Deliberate targeting "prosecutes planned targets... that are known to exist in the operational environment with engagement actions scheduled against them." Dynamic targeting "prosecutes targets of opportunity... identified too late, or not selected for action in time, to be included in deliberate targeting" and also manages changes to planned targets when the situation shifts.
Both categories run through the same six-phase joint targeting cycle - (1) end state and commander's objectives, (2) target development and prioritization, (3) capabilities analysis, (4) commander's decision and force assignment, (5) mission planning and force execution, (6) assessment - and they "differ only in the execution of phase 5." Deliberate targeting works targets methodically ahead of time; dynamic targeting runs the F2T2EA steps live in phase 5, because the decision to engage has to be made quickly. A frequent application is the time-sensitive target: one "of such high importance... or one that presents such a significant strategic or operational threat" that the commander "dedicates intelligence collection and attack assets" to prosecute it, concentrating the whole chain's difficulty into a very short window.
Because every stage before engage is a question about data, and the answer has to be assembled from many sources that rarely agree on their own. Finding a target means separating a real detection from noise across traditional and nontraditional sensors. Fixing it means resolving one trusted location out of reports that lag, conflict, or use different coordinate frames. Tracking it means stitching intermittent observations into one continuous history as the target moves. Targeting it means fusing that track with identity, context, rules of engagement, and risk. None of those are weapons problems. They are correlation problems, and JP 3-60 says as much when it calls find, fix, and track the intelligence-intensive part of the chain.
The difficulty has grown because the underlying data is fragmented. A single target's signature can be spread across a radar return, a full-motion-video feed, a signals report, and an open-source cue, each in a different system, format, and security domain, each arriving at a different time. No one source closes the chain. The target only becomes actionable in the correlation across sources, and doing that correlation at the speed a moving target demands is a data-infrastructure problem. This is the same cross-source reasoning that multi-INT fusion performs for the analyst and that AI-enabled ISR and PED accelerates at the processing, exploitation, and dissemination step. It is also why a trustworthy common operating picture is a prerequisite for a fast chain: if find, fix, and track draw from an inconsistent picture, every downstream decision inherits that uncertainty.
AI compresses the kill chain by collapsing the time spent in its data-intensive stages, not by automating the decision to engage. The slow part of F2T2EA is almost never the physics of the engagement; it is the correlation work in find, fix, and track, and the exploitation that turns raw collection into a usable track. When that fusion happens at machine speed - resolving when separate reports describe the same object, maintaining a track across sensors, and surfacing the identity and context a decision needs - the whole chain tightens because orientation gets faster. This is the targeting-level expression of what CJADC2 is meant to deliver across the joint force, and AI's part in it is a CJADC2 data and reasoning role: connect the sources, fuse them into shared understanding, and route that understanding to where a decision is made.
Two cautions keep this honest. First, speed is only an advantage if the picture is trustworthy; a chain that closes fast on a mis-correlated track is worse than a slow one. Compressing the chain therefore depends on fusion that preserves provenance, so a human can see why two reports were judged to be the same target and defend the resulting decision. Second, doctrine keeps a human in the loop for the target and engage stages by design. AI's role is to compress the data stages - find, fix, track, and the exploitation feeding them - so the people making the engage decision have more time and a better-grounded picture, not less control. The goal is a faster, more defensible decision, with human judgment preserved where doctrine requires it.
Torch.AI does not build weapons or a targeting application, and this page is not a claim to one. What Torch.AI builds is the government-owned reasoning infrastructure the data-intensive stages of the kill chain depend on: connecting fragmented, multi-source data into one coherent, trustworthy picture at machine speed while keeping every element traceable to its source.
The relevant capability is fusion you can trust and govern. Closing a chain faster means pulling detections and reports out of many systems, resolving when separate records describe the same object, maintaining that resolution as a track, and carrying identity and context into the decision - all without trapping mission data in a vendor's proprietary model. ORCUS provides the data-availability layer that connects to sources and preserves provenance. NEXUS applies semantic representation so unstructured records become something a system can correlate. HALO fuses resolved entities and relationships into a graph an analyst can inspect and trace. Because this runs as a system of reason on top of existing systems of record, the authoritative sources stay intact, and in a government-owned deployment the customer keeps control of the layer where fragmented data becomes understanding. See how this is packaged on the Torch.AI software page.
For evaluators, the fit is specific: if what slows your chain is that detections, tracks, and context are scattered across systems faster than anyone can correlate them, that is a data and reasoning problem, and it is the one this infrastructure is built for. Request a technical walkthrough and we will run it against a representative slice of your own multi-source data, with provenance traced end to end.
Targeting doctrine is settled. The modern gap is the data capability underneath the chain. If you are evaluating infrastructure to shorten the time from detection to decision, require:
Evaluating a capability? Bring these six requirements to a scoping call and we will walk each one against your environment: request a technical walkthrough.
The kill chain has six stages: find (detect and classify a possible target), fix (determine its location), track (observe its activity and movement), target (decide how to engage and coordinate the means), engage (take action against it), and assess (examine the results). In JP 3-60 the first five are phase 5 of the joint targeting cycle and assess is the cycle's final phase. The steps can run iteratively and in parallel rather than in a strict line.
The kill chain is a specific targeting workflow (find, fix, track, target, engage, assess), while the OODA loop (observe, orient, decide, act) is a general decision-making model developed by John Boyd. They are complementary: the kill chain can be seen as OODA applied to targeting, with find and fix mapping to observe and orient, target to decide, and engage to act. Both reward faster, better-oriented decisions built on well-fused data.
Dynamic targeting is the prosecution of targets of opportunity that emerge too late to be planned in advance. It is phase 5 of the joint targeting cycle, run using the F2T2EA kill chain, and it contrasts with deliberate targeting, which prosecutes planned targets known in advance. A common application is the time-sensitive target, a high-importance or high-threat target the commander dedicates assets to prosecute quickly.
Because every stage before engage depends on fusing fragmented multi-source data into a trusted picture. Finding, fixing, and tracking a target are intelligence- and sensor-intensive, and a target's signature is usually spread across different systems, formats, and security domains. The target only becomes actionable in the correlation across those sources, so the chain closes only as fast as that correlation can be done reliably.
AI compresses the kill chain by speeding up its data-intensive stages - find, fix, track, and the exploitation that turns raw collection into a usable track - so orientation happens faster and the whole chain tightens. It does this by correlating multi-source data at machine speed while preserving provenance. It does not and should not automate the engage decision; doctrine keeps a human in the loop, and AI's role is to give that human more time and a better-grounded picture.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
