Positiv Cohort helped a defence organisation progress from an initial AI proof of concept towards a more capable multi-agent solution – establishing a practical foundation for AI-assisted engineering within its secure, air-gapped environment.
The Challenge
A defence organisation was exploring how artificial intelligence could support the creation and review of complex engineering requirements within an advanced defence programme. Requirements engineers work with large volumes of specialist information and must ensure that individual requirements remain consistent with wider designs and organisational knowledge.
Supporting these checks with AI could reduce manual effort, help engineers work more effectively under time pressure and contribute to faster development cycles.
The environment made innovation particularly challenging. The work had to be conducted within secure, air-gapped infrastructure where access to technology and computing resources was constrained. Processes designed to protect sensitive information also made collaboration and technical progress more complex.
The organisation needed a partner that could combine specialist AI knowledge with an understanding of the security, assurance and systems engineering disciplines required within advanced defence programmes.
The Positiv Cohort Solution
Positiv Cohort worked in partnership with the organisation to design, develop and adapt an AI platform around its engineering requirements and secure operating environment.
We began by developing a proof of concept using retrieval-augmented generation, or RAG. The solution compared engineering requirements with wider design information and relevant organisational knowledge to support consistency checking.
Requirements engineers who evaluated the proof of concept found that it produced useful results and could assist them during time-pressured reviews. Testing indicated performance in the region of 75–80%.
The initial work also revealed the limitations of asking a single AI model to interpret different forms of specialist information and complete several distinct tasks. Rather than treating this as the limit of the opportunity, we used the findings to shape the next phase of the solution.
“Positiv Cohort’s team brought the secure agility and speed needed to innovate within a highly constrained environment. When one route was unavailable, they found another compliant way to keep the work moving.”
Data & AI Engineer | Secure Defence Programme
Developing a Specialist Multi-Agent Solution
To address the limitations identified during the initial proof of concept, we progressed the platform towards an agentic RAG architecture.
Instead of relying on one model to perform every task, the solution assigns specialist responsibilities to different AI agents. These can include interpreting design information, applying engineering knowledge, producing written outputs and evaluating their quality.
Evaluator agents assess outputs against defined criteria and can return work to the specialist agents for further refinement. This creates an iterative process designed to improve accuracy and completeness before the output is presented to the user.
A working version of the multi-agent solution has been deployed within the client’s secure, air-gapped environment. The next stage is to validate whether the architecture can overcome the performance limitations encountered during the original proof of concept.
Maintaining Progress Within a Secure Environment
The programme required more than AI engineering and architecture expertise. Progress depended on persistence, experience and an ability to work constructively across organisational boundaries.
For Positiv Cohort, persistence meant treating each obstacle as a problem to be understood and resolved, rather than a reason for the programme to stall. When one route was unavailable, we explored other compliant options, engaged the appropriate departments and built the relationships needed to move forward.
Security requirements meant that different teams had valid reasons for operating independently. We worked across those boundaries, shared appropriate knowledge and supported other specialists while continuing to respect need-to-know principles.
This included mentoring client engineers, combining practical experience with academic rigour, and helping the organisation consider both where AI could add value and where it should not be used.
For the client, Positiv’s approach maintained momentum without bypassing security or governance. It helped move the programme from an experimental concept to a working solution within an environment where innovation can otherwise take considerably longer.
Programme Progress
The engagement has so far delivered:
- a working RAG proof of concept for engineering consistency checks;
- practical evaluation and feedback from requirements engineers;
- a clear understanding of the initial approach’s performance limitations;
- a specialist multi-agent architecture informed by those findings;
- a working version deployed within the secure environment;
- knowledge sharing and mentoring for the client’s engineers;
- a foundation for exploring further responsible applications of AI.
The longer-term ambition is to use AI to reduce the effort associated with engineering requirements and contribute to shorter development cycles.
Defence Pedigree in Practice
Complex platforms and autonomous systems programmes depend on the interaction of engineering disciplines, technologies, assurance processes and operational requirements.
Our work combined AI engineering and architecture expertise with the persistence and judgement needed to operate effectively within a secure defence environment. We adapted the technical approach in response to evidence, worked across organisational boundaries and transferred knowledge throughout the engagement.
This gave the client a professional services partner capable of moving beyond an isolated technology demonstration and progressing the solution around its real engineering needs, infrastructure and constraints.
How We Support Platforms & Autonomous Systems Programmes
This engagement demonstrates how Positiv Cohort can help defence organisations apply emerging technology to the complex engineering processes that underpin advanced platforms and systems.
Our wider Platforms & Autonomous Systems pedigree spans the Type 26 Frigate, military vehicles, UxVs and Counter-UxV systems. We bring together specialists who understand the relationships between platforms, autonomous technologies, systems engineering, assurance and the wider operational environment.
We work in partnership with client teams to understand the requirement, develop proportionate solutions and maintain progress through complex technical and organisational constraints.
Whether improving engineering processes, integrating emerging technology, developing autonomous capabilities or supporting platforms through their lifecycle, we help clients turn demanding requirements into practical progress – without compromising the security, governance and engineering discipline that defence programmes demand.