Closing the Feedback Loop: Data Exchange Between AI-Targeting System Developers and Combat Units

The advancement of AI-enabled targeting systems is constrained not by technology but by a broken feedback loop. Data exchange between developers and front-line crews remains informal and fragmented: without a unified data-collection protocol, manufacturers cannot iterate reliably and crews are overburdened by duplicated requests — slowing the adaptation of weapons to a rapidly changing battlefield.

The Operational Challenge

Communication between developers and crews is situational and flows in two informal directions: developers reach out to crews to query a system's performance, and crews report malfunctions or errors back to developers. Each manufacturer builds its own approach to data collection — defining its own formats, marking requirements and transmission channels. The absence of a common standard fragments these processes and makes coordinated work between the parties difficult.

The current model places a significant burden on service members, offloading part of the data-collection effort onto the crews themselves. This feeds a persistent perception that the testing of immature solutions consumes a military resource for commercial benefit. Where feedback yields no visible improvement, crews' motivation to cooperate erodes — and with it, the quality and continuity of the data on which the systems depend.

Analytical Assessment

The assessment is based on qualitative research: fifteen semi-structured in-depth interviews (N=15) with front-line operators of strike and reconnaissance UAS, developers of AI-enabled targeting modules, a representative of a national defence-tech platform, and a subject-matter expert. Data was collected in February–March 2026 and analysed in March–April 2026. The sample was purposive and self-selecting; findings therefore reflect the positions of individual participants within this problem space.

Three root causes emerged.

1. No agreed protocol of interaction. Service members have no clear list of the data required, nor of how often to submit it. There is no shared understanding of what to collect, how to mark it, or through which channel to send it.

2. Insufficient coordination among developers regarding crew needs and constraints. Manufacturers do not consistently account for what crews can realistically do under combat conditions. As a result, crews engage several developers in parallel and transmit the same information repeatedly, creating duplication and excess load. Technical solutions that would minimise manual crew involvement remain rare and do not scale.

3. No coordinating structure in the field. Periodic exchange events do not guarantee developers a stable flow of data, nor crews a simplified collection process. Coordination must do more than gather expectations — it must produce a single standard for collecting and transmitting information.

Conclusions and Recommendations

The current model of interaction between developers of AI-targeting systems and crews is decentralised, fragmented and built largely on informal personal contact. The absence of common standards and transmission protocols overburdens service members, forces developers to duplicate effort, and slows the adaptation of technical solutions to a fast-changing operational environment.

Overcoming these barriers requires a shift to systemic coordination: a single secure channel and a unified data-collection protocol. The priority is the automation of telemetry collection to minimise human intervention. Critically, standardisation should apply only to data formats and the user experience — it must not constrain competition or the innovation potential of developers. In a workable model, crews submit data not to individual developers but to a coordinating entity that performs primary filtering, processing and controlled-access storage. A standard validated by the end-users themselves stands a far greater chance of rapid adoption and scaling, because it reflects the real needs and constraints of those who operate the systems.