Enterprise general intelligence is execution

Enterprise general intelligence describes operational capability that transfers across functions and tenants. It is not a claim about artificial general intelligence or a general-purpose chatbot. In this architecture, generality comes from reusable execution patterns that are benchmarked before tenant deployment.

A more capable model, enterprise retrieval, and a larger context window can improve interpretation and knowledge access. They do not define transaction semantics or functional completion. Those still have to be implemented and evaluated for each operational job.

What “general” is allowed to mean here

• General across foundation models — inference is replaceable when paired job evaluations preserve completion and controls.
• General across unlike industries — the same job spec, scored at multiple tenants.
• General across functions — purchasing, finance, warehouse, operations, retail, revenue, and product on one execution layer.
• Not general as a blank-slate personality that you prompt into being a controller, a buyer, and an SDR at once.

Generality does not imply shared permissions or interchangeable workflows. A purchasing sequence must not inherit finance write paths. Function-specific tool catalogs constrain capability, the control plane constrains authority, and action-level evaluations measure whether the resulting work is correct.

Generality is measurable when the same job remains correct across models, tenants, and functions without relaxing function-specific controls.

In this formulation, generality is measured by transfer across models, tenants, and functions while preserving function-specific controls. Bruce is the production implementation of that architecture.