Capabilities
Background arrives with a library of built-in capabilities covering the most important situations in construction and infrastructure. As it learns your project — your RAMS, your programme, your contract — more emerge.
How capabilities work
A capability is not a sensor trigger or a threshold alarm. It is a pattern of evidence — observations combined with context — that CompoundOS™ recognises as a situation requiring attention. No single observation fires a capability. Evidence combinations do.
When a capability fires, it produces an OperationalObservation — a structured piece of understanding that includes what was observed, what evidence supports it, what the confidence level is, and what intervention is recommended. Every output is traceable back to the evidence that created it.
Situations where people may be exposed to harm — grounded in the RAMS, permits and supervision requirements that define what safe looks like on each work package.
Situations where programme momentum is at risk — connected to the programme, workforce data and site activity to identify constraints before float is consumed.
Situations where commercial position is at risk — connected to the contract, programme and site observations to surface obligations before they expire.
Situations where quality standards are at risk — connected to inspection records, method statements and the quality requirements defined for each work package.
Emerging capabilities
The capabilities above are built in — available from day one. But Background also learns from each project it operates on. As CompoundOS™ accumulates evidence from your RAMS, your programme and your contract, it identifies patterns that are specific to your project, your organisation and your environment.
These emerging capabilities are not generic. They reflect the specific situations that matter on your site — situations that no pre-built library could anticipate, because they emerge from the intersection of your particular context and the evidence that Background observes.
The defensible position
Background's capabilities are not dependent on any single AI model. The underlying language model can be replaced or upgraded. What cannot be easily replicated is the Construction World Model — the ontology that makes observations meaningful, the capability library that defines what matters, and the accumulated evidence relationships that improve with every project.
This is what makes Background defensible over time. Not the model. The understanding.
Next step
Walk through a real scenario — capability firing, evidence traced, understanding delivered, intervention created. From observation to outcome.