Controlled AI Execution

Every AI task should earn its execution.

Information crosses on your terms.

Kordane decides whether a sensitive task may run, where it may run and which model is sufficient. It records the policy, evaluation evidence, fallback path and final decision.

Controlled AI for sensitive workflows: private cloud, on-premises or isolated infrastructure, with frontier models only when the task genuinely requires them.

Discuss one workflow See a controlled crossing

Private by architecture
Policy enforced
Auditable by design
Deploy where required

What you control

Control the crossing, the execution and the cost.

Control sensitive information
Detect, classify, transform, or block restricted information before it reaches an AI system.
Control execution
Run each task only in a permitted environment and preserve the decision that allowed it.
Control model cost
Use the smallest permitted model capable of meeting the required quality threshold.

Interactive demonstration

Follow one controlled crossing.

See how a request is classified, governed, transformed, routed, executed, restored, and recorded. A user or application takes restricted material to an AI system, and a response or an action comes back. Nothing decides what may leave, which model may see it, or what record remains. Kordane inserts a policy-controlled boundary into that path.

User or applicationSensitive contextAI model or toolResponse or action

Policy profile

Sensitive values are tokenised or redacted, codewords may not leave at all, and the request is routed only to an approved external model.

Synthetic request

Synthetic security demonstration. Detection, policy, and transformation run locally in your browser. The model response is generated deterministically for illustration. Nothing is transmitted.

Ready. Choose a profile and run the crossing.

Original requeststays local

Sanitised outbound requestwhat would cross

Policy & audit timelinereviewable trace

Route: the approved external model.

authorisedtransformedblockedcontained

Synthetic model responsewaiting

Restore is a local substitution inside your browser. Reset clears the request, result, and audit trail.

The decision record

An audit log records what happened. A decision record explains why it was allowed.

Every controlled execution closes with one structured record: the policy that permitted it, the environments that were valid, the models that were considered and the evidence retained. This is a synthetic example.

Kordane Boundary · Decision recordKDR-2026-0714-00381

Summarisation of a restricted investigation report was permitted in the private VPC using the specialised private model, which cleared the 0.92 quality threshold. Protected identifiers were tokenised locally before the crossing. The approved frontier route is held as fallback only.

Task: Summarise a restricted investigation report
Information class: Restricted
Requested by: Named analyst with authorised case access
Policy version: boundary-policy-17
Permitted environments: Private VPC · isolated local
Rejected environment: Public frontier API, refused for this information class
Quality threshold: F1 at or above 0.92 on the case-summary evaluation set
Transformation: Protected identifiers tokenised locally
Fallback condition: Candidate evaluation falls below 0.92 before deployment or during scheduled re-evaluation; escalation runs through the approved frontier route, after transformation
Human approval: Not required for summarisation · required before any external action

Candidate models

local-small-v3rejectedF1 0.89 · Below the 0.92 threshold
specialised-private-v2selectedF1 0.94 · Smallest permitted model above threshold
approved-frontierfallbackF1 0.96 · Held as fallback; not required

Evidence retained

Policy decision
Information classification
Transformation record
Evaluation set version
Candidate comparison
Selection reason
Fallback path
Response provenance
Decision identifier

Synthetic example · Illustrative decision record · Not customer evidence · Not a measured production benchmark

See the full evidence model

Control platform

Two decisions before any task runs.

Boundary decides whether and where a task may run. Forge decides which permitted model should perform it. Optimisation never overrides policy.

BoundaryProtect · Policy · Route · AuditBoundary controls permission and environment.ForgeEvaluate · Select · Train · OptimiseForge controls model sufficiency and efficiency.

Not a gateway. Not a masking proxy. Not just a private cloud. The difference is the decision. How Kordane compares

Applications built on the platform

CortexPrivate assistants, document intelligence and speech, grounded in your approved sources.Explore Cortex →
AxonCustomer conversation automation with a governed human handoff.Explore Axon →

Integration between products is configured per workflow and deployment; each product carries its own maturity, stated on its page.

Explore the product portfolio

Evidence

See the products at work.

Illustrative product views · Synthetic data

Deployment and trust

Runs inside your constraints. Proves what it did.

The same boundary architecture is designed to deploy from approved cloud to fully air-gapped environments, on infrastructure you control. No badges, no absolutes: mechanisms are explained, each control carries its status and limitations are documented.

Where it runs

Trust boundaries and controls

The products consolidate working systems the founders have already built; Kordane is opening its first design-partner pilots. The public demonstration is synthetic: deterministic, local logic on fictional data. Production controls are scoped per pilot, and security documentation is shared during qualified pilot discussions.

Deployment patterns and matrix →Trust architecture and threat model →

Bring one sensitive workflow.

We will map the information crossing, identify the required controls, recommend an execution pattern, and determine whether a focused pilot is justified.

Discuss one workflow What a pilot includes