Carraggon One — One World Models
One model of your world, not one more chatbot.
Carraggon One is a family of world models — Spatial & Physical, People, Events, Enterprises and Healthcare — that learn how your organisation and its environment actually behave, then let you simulate what happens next before you commit to it.
World models
Five, one substrate
Horizon
Seconds to quarters
Grounding
Your data, your rules
Output
Simulate, then decide
Language models know what people have written. World models know how things move, who does the work, what happened before, and what breaks if you change it.
The family
Five world models. One world state.
Each One WM specialises in a different layer of reality, and all five write into the same representation — so a delayed shipment, a short-staffed shift and a missed revenue target are understood as one story.
01
One Spatial & Physical WM
Space, assets and motion — modelled as a place, not a spreadsheet.
Buildings, floors, sites, fleets, machines and the people moving through them, held in a continuously updated geometric and physical state. Ask what happens if a line goes down, a wing is closed, or demand doubles, and get a grounded answer.
- —Digital twins of sites & assets
- —Occupancy, flow & routing
- —Sensor and IoT fusion
- —Physics-aware what-if simulation
02
One People WM
How work actually moves through the humans who do it.
Roles, skills, teams, workloads, handoffs and tenure — modelled as a living organisational graph. Forecast capacity, attrition risk and the second-order effect of a reorg before you commit to it.
- —Skills & capacity graph
- —Workload and burnout signals
- —Attrition and succession forecasts
- —Reorg and hiring simulation
03
One Events WM
Every signal on one timeline, with cause kept attached to effect.
Tickets, transactions, incidents, deployments, calls, claims and market moves ordered into a single causal event stream. Detect the pattern early, replay the sequence, and test the intervention against history.
- —Unified causal event stream
- —Early anomaly detection
- —Root-cause replay
- —Counterfactual intervention tests
04
One Enterprises WM
The company itself as a model you can run.
Entities, contracts, cost centres, suppliers, customers and cash — connected so a change in one shows up everywhere it matters. Model pricing, expansion, supplier failure or a new market against your real operating structure.
- —Multi-entity operating graph
- —Contract & supplier exposure
- —Revenue and cost propagation
- —Scenario planning to P&L
05
One Healthcare WM
Patients, pathways, capacity and coverage in one clinical model.
Care pathways, staffing, beds, referrals, claims and outcomes represented together, with privacy enforced at the boundary. Plan capacity, spot deterioration earlier and test protocol changes safely.
- —Care pathway modelling
- —Capacity & staffing forecasts
- —Risk stratification
- —Claims and coverage linkage
What makes it a world model.
One representation
Each model writes into the same world state, so a person, a place, an event and an entity are the same objects everywhere.
Grounded, not guessed
World models are trained and continuously corrected against your systems of record — not a general model's recollection of the internet.
Simulation first
Every recommendation comes with the run that produced it: assumptions, horizon, confidence and the counterfactual it beat.
Governed by default
Residency, redaction, role-scoped access and full audit of every read, run and write.
Capabilities
Predict, explain, simulate, act.
Predict
What happens next, with a stated horizon and confidence band.
Explain
Which prior events and dependencies drove the outcome.
Simulate
What changes if you intervene — before you intervene.
Act
Hand the chosen plan to Carraggon agents to execute and monitor.
How it works
From raw systems to governed decisions.
01
Connect the world
Ingest your systems of record, sensors, calendars, ledgers and event streams into the shared world state.
02
Fit the models
Each One WM learns your specific geometry, org, cadence and economics rather than a generic prior.
03
Run scenarios
Ask questions in plain language and get a simulated run with assumptions and confidence attached.
04
Close the loop
Approved plans execute through Carraggon agents, and the outcome feeds straight back into the model.
Bring one decision. We'll model the world around it.
Pick a decision you make every quarter on instinct. We'll stand up the relevant One WM against your data and show you the simulated run behind it.