Identifying AI

Purpose

This guidance is for NSW Government public servants involved in the lifecycle of digital solutions, from design and sourcing to procurement, selection and management, who need to understand whether a solution may contain artificial intelligence (AI).

It provides general guidance only. It is not an exhaustive source for determining whether AI is present or whether an AIAF assessment is required.

Building confidence in identifying AI helps agencies recognise risks and meet their responsibilities under NSW Government assurance frameworks, including the NSW AI Assessment Framework (AIAF).

info

Register AI use cases in the AIAF Platform

From 30 September 2026, the AIAF Platform is the authoritative channel for AI use case registration and assessment across NSW Government.

All agencies must register their AI use cases. The platform will determine whether an AIAF assessment is required.

How to use this guidance

Start by reviewing the “What is AI?” section and the common characteristics of AI. Then use the questionnaire to help identify whether your digital solution uses AI.

This guidance does not replace platform registration. Agencies remain responsible for identifying AI use and maintaining appropriate agency records.

Outcome	Action
The solution uses or may use AI	Register the use case in the AIAF Platform. The registration process will determine whether an AIAF assessment is required.
The solution does not use AI	Document the decision in accordance with your agency’s governance and records-management requirements. Review the decision if the solution’s features or use change.

What is AI?

“Artificial Intelligence (AI) is the ability of a computer system to perform tasks that would normally require human intelligence, such as learning, reasoning, and making decisions.”

Source: NSW AI Assessment Framework

There is no single universally accepted technical or legal definition of AI. It can therefore be difficult to determine objectively whether a system uses AI.

Where there is a possibility that a system uses AI, register the use case in the AIAF Platform. The registration process will determine whether an AIAF assessment is required.

Common characteristics

Not all AI systems demonstrate the same characteristics. The following comparisons can help distinguish AI-enabled systems from conventional software.

Probabilistic versus deterministic

AI systems may make decisions under uncertainty and are often data-driven. Traditional systems follow predetermined rules and produce the same result for identical inputs.

Learning versus preprogrammed behaviour

Some AI systems learn or improve through data or feedback. Traditional systems generally need manual updates to change their behaviour.

Data-driven versus rule-based

AI systems use data to make predictions, generate content or support decisions. Rule-based systems execute predefined logic and workflows without learning or adapting.

Complex pattern recognition versus simple logic

AI systems can recognise complex patterns in data. Traditional systems perform tasks using explicit, straightforward logic.

Dynamic decision-making versus static workflows

AI systems may adapt decisions or outputs in response to new data and context. Traditional systems follow fixed workflows that do not change unless they are manually updated.

Questionnaire

Identifying AI can be complex, particularly when AI is embedded as one feature within a larger product. This questionnaire provides a starting point, but seek professional advice within your agency where necessary.

Start with the non-AI use cases. If the system does not meet all those criteria, continue to the AI use cases.

Non-AI use cases

Your system is unlikely to use AI if you answer yes to every question below.

Question	Example
1. Does the system operate on predefined rules without learning or adapting from data?	A basic FAQ chatbot that always provides the same response to specific questions and does not change or learn over time.
2. Does the system only store and retrieve data without analysing patterns, making predictions or generating insights?	A customer database that stores and retrieves information without producing new analysis or insights.
3. Does the system always produce the same result for the same input?	A payroll system that calculates salaries using fixed formulas and produces the same result for identical inputs.
4. Does the system perform tasks that do not require interpreting complex data formats such as text, images or audio?	A file management system that stores and organises files without analysing their content.
5. Does the system follow predefined workflows without automatically adjusting to context or data?	An automated email responder that sends the same fixed-template response to every enquiry.
6. Does the system only allow interaction through standard forms, menus and workflows, without natural-language questions or commands?	An online form that collects information through fixed steps without conversational capabilities.
7. Does the system require manual updates to change behaviour rather than adapting automatically?	Accounting software that requires a manual update when tax laws change.
8. Does the system handle straightforward tasks without pattern recognition, prediction or complex analysis?	An electronic voting system that records votes but does not analyse patterns or predict outcomes.

AI use cases

If you answer yes to any question below, the system is likely to use AI. The capability should go beyond simple rule-based processing or conventional statistical calculation.

Question	Example
1. Does the system make predictions, suggestions or decisions from data without every outcome being explicitly programmed, including through machine learning or deep learning?	A fraud detection system that identifies suspicious transactions from historical patterns and refines its predictions over time.
2. Can the system’s performance or accuracy change over time without manual rule updates?	A benefits eligibility system that learns from past cases and refines how it identifies applications requiring review.
3. Can the system analyse, interpret or generate content from complex formats such as text, images, audio or documents?	A system that transcribes recorded speech and identifies information or patterns in the resulting text.
4. Does the system make decisions or recommendations using dynamic learning or complex pattern recognition rather than fixed rules alone?	A traffic management system that changes signal timing in response to traffic patterns and current conditions.
5. Can the system engage in human-like interaction through voice or text and adapt its responses to the user or context?	A conversational assistant that interprets natural-language questions and generates responses based on context.
6. Can the system adapt its responses or recommendations using user interactions, feedback or new data?	A government portal that recommends services using prior interactions and changes its recommendations as more data becomes available.
7. Can the system perform tasks that ordinarily require human judgement, reasoning or decision-making?	A healthcare decision-support system that analyses large datasets to assist clinicians with diagnosis or treatment decisions.

What’s next?

If the system uses or may use AI

Register the use case in the AIAF Platform. The platform will use the registration information to determine whether an AIAF assessment is required. Read about the NSW AI Assessment Framework and platform process.

Use the AIAF Platform

If your agency cannot access the platform

Where SectorLink onboarding has not been completed, use the controlled offline assessment pathway described on the NSW AI Assessment Framework page. This pathway is not a general alternative to the platform. To request an offline assessment, contact NSW GovAI using the button below.

Request offline assessment

If you are still unsure

Seek advice from your agency Chief Information Officer, Digital Officer, Data Officer or other appropriate governance lead.

You can also contact NSW GovAI at govai@customerservice.nsw.gov.au.