Core technology can become a material arrangement.
APRA CPS 230 requires regulated entities to manage critical operations and risks arising from material service providers.
APRA · CPS 230 ↗Financial services & insurance / Industry brief 01
Cooma AI is developing customer-specific Australian AI estates for financial workloads where operational resilience, provider accountability and assured capacity matter alongside model performance.
The decision / Why now
The case for a dedicated estate is not “cloud versus sovereign.” It is whether a selected workload needs a more explicit operating boundary, capacity plan and service-provider agreement than ordinary consumption provides.
APRA CPS 230 requires regulated entities to manage critical operations and risks arising from material service providers.
APRA · CPS 230 ↗ASIC found 60% of the 23 reviewed financial-services and credit licensees intended to increase AI use and warned of a widening governance gap.
ASIC · REP 798 ↗ASD guidance asks cloud buyers to consider provider locality, ownership, administration, support locations and exposure to foreign control.
ASD · Cloud guidance ↗Workload selector
The operating boundary should be proportionate to the data, decision and continuity risk. Select a starting workload to inspect the recommended profile.
01 / Decision intelligence
Bring high-volume transaction and investigative workloads into a defined environment with traceable model versions, controlled data paths and planned inference capacity.
The sovereignty ledger
The ledger updates with the selected workload. It is the starting point for a control conversation—not a claim that infrastructure alone creates compliance.
Keep transaction and investigation data on approved Australian processing paths with explicit retention and deletion rules.
Version, evaluate and approve the models that generate alerts, scores and investigator support.
Define Australian privileged administration, screening, access approval and separation of duties.
Agree logging, key ownership, incident handling, model change control and the provider chain.
Plan inference headroom, failover, recovery priorities and exit before the workload becomes critical.
Proposed deployment pattern
Cooma's intended role is to integrate with the institution's existing data and cloud environment, while making the selected AI workload's operating boundary and capacity explicit.
Systems of record, governed data products and existing integration services.
Workload profile, service boundary, capacity envelope and procurement requirements.
Customer-specific infrastructure and operations subject to executed customer and supply agreements.
Approved APIs and services integrated into business processes with accountable human oversight.
Cooma is developing this operating model. Final architecture, security, capacity, availability and delivery terms depend on the workload and executed customer, facility, power, hardware and partner agreements.
Commercial path
An MOU may record intent. The commercial goal is a sequence of increasingly specific, binding decisions.
Select one candidate workload, sponsor and target horizon.
Qualified opportunityConfirm data class, service criticality, current environment and decision owners.
Mutual NDA + scopeProduce the operating boundary, architecture, demand envelope, risk map and procurement brief.
Binding paid SOW · 4–6 weeksComplete technical validation, assurance work and the production business case.
Paid milestonesContract defined service, acceptance, minimum commitment and expansion terms.
Long-term service revenueFinancial services briefing
Do not send customer, transaction or otherwise sensitive information. The first discussion needs only the business outcome, workload type and target horizon.
Evidence & status
Research updated 23 July 2026
Primary context: APRA CPS 230 · APRA CPG 230 · ASIC REP 798 · ASD cloud guidance. Requirements depend on the entity, workload and arrangement. Cooma's estate and service descriptions are proposed and subject to contracting and delivery readiness.