Kode-1
Resources · Glossary

The vocabulary of governed AI.

The terms we use when we talk about AI in production, and the standards we work to, defined for engineers, executives and auditors alike.

How the work is done

Forward-deployed engineering
An embedded delivery model: senior engineers work inside the client's environment, alongside its own teams, to take approved use cases into governed production. The client's security model, legacy systems and compliance obligations are design inputs from the first week, and the work ends with the capability handed over.
Forward-deployed engineering →
Governed production
The state an AI use case reaches when it has a named owner, its policy enforced on every request, every interaction traced, its cost attributed to the business that benefits, a retained record of what it did, and a scorecard someone with authority reviews on a schedule. Anything short of that is still a pilot.
AI delivery and assurance →
AI control plane
The runtime layer across an AI stack that gives one place to set policy, capture audit evidence and attribute cost for every interaction. It is what makes governance real while systems are running, rather than a set of documents.
AI security →
Shadow delivery
A way of transferring capability in which the client's team runs a live use case end to end while experienced people pair with them, coach against the method and review each artefact. The team is signed off to deliver independently once agreed readiness gates are met.
AI enablement and capability →

AI governance and risk

AI readiness
An organisation's preparedness to run AI in production, across security, governance, observability, data, engineering, operations, value, risk and compliance, and adoption. Useful readiness is measured against evidence rather than asserted in a strategy document.
AI readiness and alignment →
AI governance
The combination of written policy, decision rights and runtime enforcement that keeps AI systems within agreed boundaries. Governance that exists only on paper cannot show what a system did; governance that is enforced produces that evidence as it runs.
The AI governance operating model →
AI risk tiering
Scoring each AI use case against criteria such as impact, reversibility, autonomy and regulatory exposure, so that higher-tier use cases attract more control: more evaluation, tighter approvals and deeper evidence. Tiering lets low-risk tools move quickly without lowering the bar for the ones that matter.
AI readiness and alignment →
Runtime guardrails
Policies that execute on every AI interaction: input validation, output filtering, data-access checks and escalation triggers. They are the difference between governance you can prove and governance you can only describe.
An AI policy has to run where the model runs →
Model drift
The gradual change in an AI system's quality or behaviour over time, often with no change to its code. It can follow a shift in the inputs it sees, an update to a provider's model or a change in the data it retrieves, and it stays invisible without continuous evaluation.
Why AI pilots stall after the demo →
Prompt injection
An attack in which crafted input causes an AI system to ignore its operator's instructions and follow the attacker's instead. Defending against it takes architectural choices, such as limiting what the system can reach, and runtime controls; careful prompt wording alone does not hold.
AI security →
AI evidence chain
The runtime record of traces, decisions, approvals, data access and cost that shows an AI system behaved within policy at a given moment. With it, an audit request is answered in days rather than reconstructed over weeks.
AI evidence and audit readiness →
AI governance forum
The cross-functional body, often called an AI council, that owns AI policy, risk tiering and exception approvals. It typically brings together technology, security, risk, legal and a business owner, and it is where the difficult judgement calls are made and recorded.
AI portfolio
The full set of AI use cases an organisation is running, scoping or scaling. Managed as a portfolio rather than as separate projects, it can be prioritised against capacity, risk appetite and value, and each use case can be expanded, held or stopped on evidence.
AI readiness and alignment →

Standards and regulation

NIST AI RMF
The AI Risk Management Framework published by the US National Institute of Standards and Technology. It is voluntary and organised around four functions (govern, map, measure and manage), and it gives risk teams a shared vocabulary for AI-specific failure modes.
AI capability →
ISO 42001
ISO/IEC 42001, the international management-system standard for AI. Built on the same structure as other ISO management-system standards such as ISO 27001, it defines what an AI management system must contain, and organisations can be certified against it.
AI capability →
APRA CPS 230
APRA's prudential standard on operational risk management for banks, insurers and superannuation entities, in force since July 2025. AI systems inside critical operations fall within its requirements on tolerance levels, material service providers and business continuity.
Financial services →
APRA CPS 234
APRA's prudential standard on information security. AI systems are information assets like any other under it, so their access, data paths and controls need the same classification, testing and incident notification as the rest of the environment.
Financial services →
EU AI Act
The European Union's regulation that classifies AI systems by risk (unacceptable, high, limited and minimal) with obligations that rise with the tier. It can reach organisations outside the EU, including Australian providers, when their AI systems are placed on the EU market or their outputs are used there.

Putting these terms into practice?

Whether you are framing a board-level technology decision, scoping a platform build, or recovering a transformation that has stalled, we lead with senior judgement rather than a sales pitch. The first conversation is always free.

Contact us