01 / FUNCTIONAL MODELLING
Model how work is accomplished
Functional System Modelling reveals the activities, dependencies and conditions through which real systems produce both successful and unwanted outcomes.
Research · modelling · evidence · learning
Safety science for complex systems.
Cambrensis develops practical methods and AI-supported applications for understanding how safety-critical systems work, vary and adapt.
System
evidence
FUNCTIONS
DATA
LEARNING
DECISIONS
VARIABILITY
evidence
HealthcareAviationProcess industriesInfrastructureResearch networks
What we do
A research and innovation hub for safety science.
Cambrensis brings together systems thinking, quantitative analysis and operational evidence. We develop methods, working models and research partnerships that help people examine complex sociotechnical systems with greater clarity.
02 / QUANTITATIVE ANALYSIS
Explore variability and uncertainty
We develop transparent analyses that connect functions, operational data, assumptions and simulation to test how system behaviour changes across conditions.
03 / INCIDENT INVESTIGATION
Reconstruct events as system behaviour
Structured investigation combines evidence, functional modelling and disciplined inquiry to explain what happened, how it emerged and why existing controls did not resolve it.
04 / AI-SUPPORTED APPLICATIONS
Make advanced methods usable
We design governed AI-supported tools that help analysts assemble evidence, build models, explore scenarios and retain an auditable line from source to conclusion.
An evidence-led approach
From system definition to useful action.
FrameDefine the system, purpose, boundary and questions.
ModelRepresent functions, couplings and operating conditions.
ConnectLink documents, observations, data and uncertainty.
ExploreTest variability, scenarios and alternative explanations.
LearnTranslate findings into decisions, trials and further research.
Application domains
Where complexity has consequences.
Our work concentrates on settings in which performance emerges from interactions between people, technology, organisation, information and changing operational conditions.
Care systems
Healthcare
Patient flow, bed occupancy, clinical work, incident investigation and the resilience of services under pressure.
Operational data
Aviation & flight analysis
Functional interpretation of flight data, actual-versus-expected performance and evidence-led examination of operational events.
Major hazards
Process industries
Functional HAZOP, barrier performance, plant operations and quantitative exploration of how deviations propagate through complex systems.
Under controlled evaluation
FRAIXL™
Functional modelling, evidence and governed AI reasoning
An emerging workbench for functional modelling, evidence and governed systemic reasoning.
FRAIXL™ is an emerging workbench for understanding and analysing complex sociotechnical systems. It combines executable Functional System Modelling, evidence, quantitative analysis and simulation with SAGE 1 — Systemic Agent Governance & Epistemics, our protocol for governed agentic reasoning.
Rather than asking an AI to absorb an entire system and produce an answer, FRAIXL keeps the model, evidence and system state explicit. SAGE 1 then governs how AI reasons over selected parts of that system.
- Functional System Modelling
- Evidence & provenance
- Executable analysis
- SAGE 1 reasoning
- Challenger analysis
- System Synthesis
- Human review
Selected supervised trials are being used to establish how well the workbench supports users who were not involved in its development.
Discuss a supervised research trial FRAIXL is a trade mark applied for by Cambrensis Ltd. UK application UK00004446658.Governed AI reasoning
SAGE 1
Systemic Agent Governance & Epistemics
AI that reasons within the system, not instead of it
SAGE 1 is the governance protocol used for agentic reasoning in FRAIXL and IRAG3. It constrains reasoning to evidence-linked, graph-bounded questions while preserving the wider system context.
Analyse locally. Explain systemically.
AI reasoning is organised into bounded Analysis Atoms, each tied to evidence, system state and surrounding relationships. Conclusions receive an explicit epistemic status so that evidence, derivation, inference, assumptions, hypotheses, contested interpretations and unresolved questions remain distinguishable.
Deterministic checks are applied wherever possible before AI interpretation is accepted. A separate Challenger tests assumptions, contradictory evidence and alternative explanations, while risk-based review directs human attention towards the findings that matter most.
System Synthesis then brings local analyses back together to examine feedback, interacting variability, shared constraints, adaptation, barrier dependencies and other behaviour visible only at system level. This prevents detailed AI analysis from becoming reductionist.
From evidence to governed human review
- Evidence
- System / Incident Graph
- Analysis Atoms
- Mechanical Validation
- Epistemic Classification
- Challenger
- System Synthesis
- Human Review
Model the system. Govern the reasoning. Understand the whole.
Analyse locally. Explain systemically.
The Analysis Atom is the unit of reasoning, not the unit of system meaning.
Whole-system context → local analysis → interaction between local findings → System Synthesis → revised understanding of the whole.
Current research status
Implemented and under controlled research evaluation- SAGE 1 has been implemented as an additive research capability.
- 1,227 unique checks have been verified.
- Demonstrations have been completed in Cement, Hospital, Air / Stable Approach, and IRAG3 incident analysis.
- Analysis Atoms, controlled epistemic status, bounded retrieval, Challenger analysis, adjudication, review priority, replay and cross-atom System Synthesis have been implemented.
- The generic FRAIXL kernel remains unchanged.
- Domain models remain unchanged.
- System-level interpretations remain human-confirmed.
- The capability is not yet deployed as an independently qualified operational system.
Governed roles for software, AI and people
SAGE 1 keeps each contribution explicit so that AI supports expert judgement without replacing it.
System modelProvides structure and preserves the wider operating context.
EvidenceProvides provenance and a traceable basis for claims.
Deterministic softwarePerforms checks that can be resolved mechanically.
AI reasoningContributes interpretation, pattern recognition, hypotheses and challenge.
System SynthesisExamines interaction, emergence, feedback and shared constraints.
Human authorityRetains responsibility for important system interpretations and decisions.
Collaboration
Building the work around the problem.
Cambrensis convenes the combination of operational, academic, technical and domain expertise that complex safety questions require.
Research consortia
We are exploring research consortia in three areas: healthcare systems, flight-data analysis and process industries. Each would bring operational evidence together with a functional model to examine how the system performs. We welcome interest from researchers, practitioners and organisations with relevant questions, expertise or suitable data.
Express your interest ↗International conferences and workshops
We organise focused events that bring researchers, practitioners, regulators and technology partners together to examine methods, cases and emerging applications.
Research & insights
Ideas made open for examination.
Our papers, working notes and essays explore the foundations, limitations and practical development of safety science.
Methods · Sep 2026
Functional Modelling of Complex Systems
From systems thinking, SEIPS and FRAM towards executable functional models.
Open paper → AI & risk · Jul 2026From HAZOP to Risk Intelligence
Can governed AI and explicit evidence restore confidence in quantitative risk assessment?
Open paper → Perspective · Jul 202650 Years of Safety “Science”?
A personal view of the field from someone who was there at the beginning.
Open essay →Start a conversation
Bring us the system that needs to be understood.
We welcome conversations with research partners, safety practitioners, universities, healthcare organisations, aviation specialists and process-industry teams.
info@cambrensis.org