Editorial – ITEA Journal – June 2026

It has been a pleasure to be guest editors for this issue of the ITEA Journal of Test and Evaluation. Excellence in T&E benefits from exchanging technical, programmatic, and acquisition knowledge across the community. This journal—alongside the workshops, conferences, fora, summits, and symposia—keeps us advancing towards that goal.

The core of this issue comes from posters presented during the AI in T&E Forum, held March 17 – 18, 2026 in Washington DC. The aim of the forum was to share the latest T&E policies and guidance in the context of emerging AI-enabled systems being integrated into military capabilities. The focus of the forum was the safe and effective integration and validation of these complex technologies into existing platforms. In other words, the goal of this community – making sure things work correctly for the warfighter.

Articles in this issue align with that focus and directly support the Warfighting and Enterprise mission areas of the Artificial Intelligence Strategy for the Department of War. The articles address how to strengthen decision making through T&E of AI, as well as incorporating AI into the T&E practitioner’s toolbox to modernize the daily workflow. In addition, we also include two articles which contribute to the same goals without AI. We sincerely hope you enjoy this collection of articles.

We present the following five articles based on posters presented at the AI in T&E Forum.

Tightening the link between testing and decision making, Eman Kawas has provided a briefing note expanding her poster on Decision Assurance for AI-Enabled Mission Systems: From Test Evidence to Operational Authority. Eman’s note provides a conceptual framework to link test evidence to operational decision authority. This structured method targets the gap between test evidence (traditionally focused on performance metrics) and operational authority exacerbated by adaptive AI capabilities. The Decision Assurance paradigm would be used to inform when AI-enabled capabilities could be deployed with appropriate confidence, governance, and operational oversight. Accompanying the framework is an illustrative case study of how the framework could be applied to an AI-enabled mission planning scenario.

To accelerate the workflow, Dr Muhammad Islam, Dr Tomi Esho, and Dr Jyotirmay Gadewadikar have provided their poster with an accompanying briefing note, An AI-Enabled Framework for advancing Test and Validation across the Systems Engineering Lifecycle. This introduces an AI-enabled, requirements-based test generation framework to reduce effort, speed delivery, and improve mission assurance. Their analysis shows that test cases generated using their AI4SE framework are comparable to human written test cases, when evaluated by experts across five essential criteria. These initial results appear promising, offering a reduction from hours to minutes for test case development, while retaining a human in the loop. Further expert evaluation, data collection, and beta testing are being planned.

To accelerate the assurance of that workflow, Mehran Irdmousa expanded on his poster to produce a paper on the Systems Engineering Command Center: An AI-Powered Assurance Agent for Complex Government System Integration. The agent uses retrieval-augmented generation and locally hosted large language models to be an AI-powered assurance platform at the artefact level. This method offers automated consistency checking, traceability verification, and compliance validation across all provided artefacts. Local hosting is an important intentional architectural decision for programs handling sensitive documentation. The paper also presents evaluation findings from an initial proof-of-concept implementation. Targeting internal consistency, completeness, and mutual compatibility of the defining system specifications and requirements furthers the efforts to shift test left, to identify issues earlier and cheaper.

Making sure humans keep an eye on the process, Andrew Pollner’s paper on Human Oversight for AI-Generated Test Artifacts gives us thought to truly understand the capabilities and limitations of the artificial intelligence we are using; in this case, as used to generate test artifacts. The paper does not advocate for non-use of artificial intelligence in test artifact generation but rather offers a more durable approach by combining the positives of artificial intelligence acceleration with structured oversight. The paper makes the argument that whatever you use artificial intelligence for, it should not be used to make final acceptance judgements, but rather its output should be treated as draft until independently reviewed and accepted by a human.

Supporting T&E of AI systems, Karen O’Brien has provided her paper Toward an Integrated T&E Framework for AI-enabled Systems: A Conceptual Model which proposes a nested set of evaluation questions for AI-enabled systems, in the same fashion as the ‘integrated survivability onion’ used for T&E within Systems of Systems. Karen walks us through from the core questions of ensuring the system does what we want it to do (and not what we don’t), out to questions ensuring the legal compliance of the system. Structured in this way, the T&E framework follows a logical and familiar pattern to ensure aspects of the new technology are adequately addressed without getting distracted by only the performance of the system.

In addition to articles from the AI in T&E Forum, this issue includes four other articles.

Continuing the AI in T&E theme, Sam Bright and Dr Michael Soltys have provided a paper titled Retrieval-Augmented Generation for Departmental Test and Evaluation. The paper details how RAG can accelerate T&E deliverables such as Test Plans, Test Reports, and Interim Fielding Assessments. It is not without careful consideration though. The workflow architecture has to be set up correctly, and security requirements need to be adhered to. The paper advocates to treat the constituent large language model as a replaceable component rather than a dependency—with a recent example to highlight the importance. Common to the theme of the AI in T&E technical papers, a human in the loop is considered mandatory in this application.

Following on from that article, Sam Bright, William Emeny, Alan Jaeger, Adam Larson, JP Lueck, and Dr Michael Soltys present Avoiding Vendor Lock-In in AI Procurement. The paper discusses the familiar experience of vendor lock-in, in the context of modern AI programs citing familiar patterns and novel challenges. Their modern context includes vendor lock-in of various types: Platform, Data, Model, and Expertise; and the authors warn about lock-in occurring even with open source elements through expertise dependencies and specialized implementations. The paper highlights that the most effective mitigations—across technical, commercial, operational, and strategic lines—combine contractual instruments with the selection of vendors that adopt standardized formats and interfaces.

Stepping away from AI, Dr Lester McCoy and Gary Honea have provided a paper on a Test Strategy and Architecture approach, titled Developing Winning Proposals through the Lens of Test and Evaluation. This paper discusses the Raytheon Test Strategy and Architecture process, which applies a test-driven focus from the outset of a program to improve successful execution predictability, ultimately supporting true rapid fielding of systems. This process brings the program management level closer to the testers through a common understanding of the test contribution early in a program, including the right sizing of the test effort against contract performance risk.

Lastly for this issue, Dr Ben Luther has provided his article titled Defining T&E as a Discipline. He offers a perspective for practitioners to keep a focus on the core reason for T&E and provides a domain-agnostic definition in support. Dr Luther discusses the problems created when the language of T&E is used loosely, and why it is sometimes narrowly interpreted as fault-finding or compliance. By focusing on what T&E is, the practitioner is able to maintain the scientific basis, communicate a clear understanding outside the discipline, and serve the demands of the systems engineering activity it feeds.

It was a privilege to be guest editors for this issue of the ITEA Journal of Test and Evaluation. We thank the team at the ITEA Journal for giving us this opportunity, as well as the contributing authors and peer reviewers for taking the time to share their expertise. Each contribution to this issue furthers the T&E field with theoretical and practical relevance to the test domain.

We challenge the reader to use these conceptual or empirical findings to guide future research, and in turn let the community know what works and what does not. As testers, we map the cliff edge to define the envelope of validated performance; and this only becomes more challenging for AI-enabled systems. We encourage anyone to participate in this process—sharing our expertise and experience allows us to learn from our peers and advance the T&E community together.