SOSX®: System-of-Systems for X is AI-enabled systems thinking, for any sector. Give it a broad question and it breaks it into component questions, researches the systems, actors and influences behind them, connects them into one network and analyses it, so the decision is made with the whole picture in view.
SOSX in under two minutes. Watch on YouTube.
SOSX: Seeing the Whole System
The central book of the series accompanying SOSX, by Chris Dean. It covers the discipline of systems thinking, the practice, and what agentic AI changes. This excerpt is abridged from the introduction, the chapters on the core ideas, and the account of what agentic AI changes. How to get the full book is at the end.
Read the excerpt (PDF, 22 pages)Five phases. One continuous pipeline.
SOSX is built for the kind of question that does not fit on a whiteboard: whole ecosystems of interacting systems, their feedback loops, and the behaviour that emerges when you put them together. The pipeline is run by thirteen specialised AI agents.
Scope
Start with a broad question. SOSX breaks it into component questions, then sets the mission, boundaries and depth: what to map and how deep to go.
Research
AI research agents search the academic and grey literature to identify the systems, actors and influences, then connect them into a typed, directed network and find the feedback loops. Every finding cites its source.
Parameterise
Quantify the network. Attach parameters (financial, labour, time, materials), each with its source and a confidence, handle the unit conversions, then define objectives and bounds for optimisation.
Analyse
Run what-if scenarios, SWOT analysis, risk registers, trade-off and gap analysis, and optimise over the network using Bayesian optimisation with sensitivity analysis and Monte Carlo uncertainty quantification.
Report
Generate reports and data exports with the sources behind every claim attached. Output to PDF, DOCX, XLSX with native charts, CSV, SVG, PNG and SysML v2.
Exports into nine MBSE tools and formats.
Outputs land where your teams already work. In engineering, SOSX exports into nine MBSE tools and formats, so the analysis feeds the models you already maintain instead of starting another silo.
SOSX sits above the toolchain, not instead of it. It is a unifying analytic layer over the tools engineering teams already use, and the same goes for people: it supports teams, it does not displace them.
Export, where you work
Results leave SOSX in a form your engineers can open in the modelling tools they already use, so the analysis carries into the models they maintain.
Shared representations
The Cameo, Rhapsody, Sparx EA and Modelio exports share one SysML v1.6 intermediate representation. Capella, Simulink and ArchiMate each have a dedicated one.
In live trials
Engineering is where SOSX is furthest along. It is in live trials with a large global aerospace manufacturer.
No new silo
We learned early that interoperability is a precondition of being taken seriously. Nobody wants another silo, so SOSX is built to feed the toolchain you have.
It supports the decision. The human stays in the chair.
SOSX is not there to make the call. It is there so that the person making the call has done the reading, can see the loops, and knows which assumptions the answer rests on.
Management decision-making based on rigorous systems analysis: every claim with its source, every value with its confidence, and every step open to review.
Every Claim Cites Its Source
Research agents search the academic and grey literature and validate findings against multiple sources. Every claim in the network and in the reports cites the source it came from, so a reader can check it.
Values With a Source and a Confidence
Parameters (financial, labour, time, materials) are attached to the network, each with its source and a confidence. You can see how far to trust a number before you rely on it, and override it with your own.
You Review Each Stage
Every build stage produces a document you can review and edit, and changes proposed by the AI are yours to accept or reject. The judgement stays with you.
Full Audit Trail
Every AI interaction is logged to the audit database, so a run can be reconstructed afterwards. SOSX is built against the frameworks an enterprise review tests an AI system against, including NIST AI RMF, ISO 42001 and the EU AI Act.
Deep analysis operations.
Once your network is mapped and parameterised, SOSX runs a suite of analyses over it, each with a dedicated panel, history tracking, exports and inline charts.
What-If Analysis
Explore hypothetical changes to the network with automatic cascade impact analysis. Conversational scenario exploration with document viewer, batch scenarios, and detailed reporting.
SWOT Analysis
Scored SWOT dimensions with radar charts, cross-session comparison, document enrichment via LLM, and history tracking. Export as XLSX with native charts or PDF.
Risk Register
Risk identification, assessment, and mitigation planning. Colour-coded XLSX export, contextual charts, severity mapping to intervention playbooks, and full history.
Trade-Off Analysis
Multi-objective comparison across systems or scenarios, with a multi-scenario radar overlay and the sensitivity caveats stated alongside each result.
Network Optimisation
Bayesian optimisation over the network, with sensitivity analysis to rank parameters by influence and Monte Carlo simulation to quantify uncertainty. Feedback loops are modelled as changes propagate.
Gap Analysis
Comprehensive or targeted gap analysis across any network. Filterable by domain, level, type, or custom hints. Identifies missing connections, undocumented systems, and coverage gaps.
Interactive network visualiser and report generation.
An interactive canvas for exploring, editing and presenting your system maps, paired with reporting that keeps the evidence attached to every finding.
Network Visualiser
A typed, directed network: systems, sub-systems, components and actors, joined by material, information, financial, influence and temporal connections
Feedback loops detected and classified as reinforcing or balancing, with a causal loop view, search and filters
Node Builder for adding, updating and deleting systems one at a time or in batches, editing connections and re-ordering the hierarchy
Clone & Sync: fork any system map to experiment, then selectively sync changes back with provenance tracked
Product Explorer for guided product-design exploration within a network, with report generation and Node Builder integration
Export & Reporting
Export as PNG, SVG, JSON, DOCX, PDF, XLSX (with native Excel charts), CSV, and SysML v2 (.sysml, JSON, XMI)
Interactive charts inline in documents with hover-to-export (PNG, XLSX). LLM-powered "Add Charts" enrichment
Every claim cites a catalogued source, and the full citations travel with the report so a reviewer can follow them
Optimisation reports covering methodology, parameter analysis and recommendations
An audit database that logs every AI interaction, user input and parameter change, so a run can be reviewed afterwards
Built against the frameworks an enterprise review tests.
SOSX is built against ISO 42001, ISO 27001, the EU AI Act, NIST AI RMF and the OWASP agentic top-10: the frameworks an enterprise security and procurement review holds an AI system to. It aligns with these standards and produces the governance evidence those reviews ask for.
UK
DSIT 5 Cross-Sector Principles, NCSC Secure AI Guidelines, DSIT AI Cyber Security Code of Practice, UK GDPR / DPA 2018, ICO AI Toolkit
ISO
ISO/IEC 42001 (AI Management), ISO/IEC 23894 (AI Risk), ISO/IEC 27001 (Information Security), ISO/IEC 27090 (AI Cybersecurity), ISO 9001
EU
EU AI Act (Regulation 2024/1689), aligned with the Articles 9-15 requirements for high-risk systems
International
IMDA MGF for Agentic AI v1.0, OWASP Top 10 for Agentic Applications, NIST AI RMF, CSA Singapore, GovTech ARC, OAuth/IETF RFCs
One platform, any domain.
The X in System-of-Systems for X is the point. SOSX is domain-agnostic: any sector with complex, interconnected systems, from healthcare to energy to financial services, can be mapped, analysed and optimised. The same engine runs each of the framings below.
MBSE
Systems-of-systems mapping and analysis at the concept design stage of model-based systems engineering. It is in live trials with a large global aerospace manufacturer.
Narrative Media
World building, lore development and narrative consistency for video games first, then TV, film and other storytelling: the story world as one living, computable model, supporting human writing teams.
Academia
Cross-referencing a body of research: connecting concepts, surfacing gaps and contradictions, and mapping a research programme from proposal to suggested directions.
Business Growth
Put a business at the centre and build out the ecosystem it operates in. Is this business going to work? How can it grow? And where should it pivot?
Investment
Ecosystem analysis as an investment pre-assessment tool: a first-pass filter on a business and the ecosystem around it, before the data room.
Utilities
The whole operating system (assets, catchments, customers, finances, regulation) as one living, connected model a utility can question directly.
Circular Economy
Material flows, actors, incentives and regulation modelled as one connected, analysable system, so you can see whether a circular loop will close before you commit to it.
Management
Decision support for executives, strategy teams and transformation offices: a decision framed as a system, its ecosystem built out around it, and the options tested before anything is committed.
Applied to the decarbonisation challenges that matter most.
ADViCE (the AI for Decarbonisation Virtual Centre of Excellence, a DESNZ-backed initiative) catalogues where AI can accelerate decarbonisation across the UK's high-emitting sectors. Industrial decarbonisation is a systems problem, and SOSX's systems-thinking pipeline addresses several of the challenges ADViCE rates highest for impact and AI suitability.
Implementing network flexibility
Map generation, storage, network, and demand-side assets as one system. SOSX quantifies the interdependencies, then runs optimisation across the whole network to find where flexibility measures deliver the most carbon and cost benefit.
Smart building control systems
Model a building portfolio as interconnected systems (HVAC, occupancy, tariffs, on-site generation) and run what-if scenarios to target control strategies where they cut the most energy across the estate, not just a single site.
Targeting optimal retrofit measures
Retrofit is a trade-off across fabric, heating, budget, and grid constraints. SOSX parameterises the estate and runs trade-off and scenario analysis to rank measures by whole-system impact, with cited, auditable reasoning behind every ranking.
Smart EV management
Map fleets, charging infrastructure, depot energy, and grid capacity as a system-of-systems. Optimise charging schedules and infrastructure placement against cost, carbon, and operational constraints simultaneously.
Low carbon manufacturing processes
Model production lines, energy flows, and feedstock dependencies end-to-end. Gap analysis and Monte Carlo optimisation surface the process changes that genuinely cut emissions, rather than shifting them elsewhere in the system.
Explore the full catalogue
SOSX's scope → research → parameterise → analyse → report pipeline applies across ADViCE's challenge catalogue, from agriculture to energy networks.
View the ADViCE Solution CardsFrequently Asked Questions
A system-of-systems is a collection of independent systems that interact to produce emergent behaviours not achievable by any single system alone, think healthcare ecosystems, energy grids, or supply chains. Mapping these reveals hidden dependencies, feedback loops, and leverage points that traditional analysis misses. SOSX is AI-enabled systems thinking: its AI agents discover, research and model these interconnections, with every claim citing its source.
SOSX works as a five-phase pipeline (Scope, Research, Parameterise, Analyse, Report) run by thirteen specialised AI agents. You give it a broad question and it breaks that into component questions. An agentic research team searches the academic and grey literature to identify the systems, actors and influences, connects them into a typed, directed network and finds the feedback loops. Other agents attach parameters, each with its source and a confidence, and optimise over the network.
SOSX treats your mapped network as a closed-box system and uses mathematical optimisation to find optimal parameter configurations. It supports Bayesian optimisation (finding optimal parameters via Gaussian processes), sensitivity analysis (ranking parameters by influence), and Monte Carlo simulation (uncertainty quantification). Feedback loop modelling handles the propagation of changes through the network. Outputs include before and after network snapshots and optimisation reports.
No. It supports the decision. The human stays in the chair. SOSX is there so that the person making the call has done the reading, can see the loops, and knows which assumptions the answer rests on. Every build stage produces a document you can review and edit, changes proposed by the AI are yours to accept or reject, and every AI interaction is logged to the audit database.
SOSX exports into nine MBSE tools and formats: SysML v2, Cameo Systems Modeler, IBM Rhapsody, Sparx Enterprise Architect, Eclipse Capella, Modelio, MATLAB/Simulink, ArchiMate and draw.io. It sits above the existing toolchain as an analytic layer, not as a replacement for it, so results land where your teams already work.
SOSX is built against ISO 42001, ISO 27001, the EU AI Act, NIST AI RMF and the OWASP agentic top-10, and aligns with UK (DSIT 5 Principles, NCSC, DSIT Cyber Code, GDPR), further ISO (23894, 27090, 9001) and international guidance (IMDA MGF v1.0, CSA Singapore, OAuth/IETF RFCs). That is alignment, not certification: SOSX produces the governance evidence an enterprise review asks for.
Bring us a real question from your business.
Get in touch to discuss how SOSX can help your organisation understand whole ecosystems of interacting systems, their feedback loops, and the decisions that depend on them.
AI software engineering and innovation
for ambitious organisations