"Automated intelligence" is not a defined category in the laws and guidance reviewed for this page, but automation and artificial intelligence are different things: automation means software performs a task without a person at each step, while US federal law frames AI by the outputs a machine-based system produces toward goals that people set. Canadian and New York City rules deliberately cover both, because the risk to a person comes from the decision, not from the technique behind it. Canada's federal directive on automated decision-making, for example, states that it is not limited to AI systems.
Key takeaways
- Automation is the broader idea; AI is a subset in which a model infers outputs from inputs.
- The US National AI Initiative Act definition (15 U.S.C. 9401) centers on a machine-based system whose outputs are predictions, recommendations, or decisions.
- Canada's Treasury Board directive covers rules, regression, and machine learning alike when a system assists or replaces human judgment.
- New York City's Local Law 144 uses a three-part test that includes "statistical modeling" and "data analytics", not only AI.
- A system can be automated without being AI, and still fall under rules written for automated decisions.
What US federal law means by AI
The National Artificial Intelligence Initiative Act of 2020 places its AI definition in 15 U.S.C. 9401. In that text, AI is a system run by machines that, working toward objectives people have chosen, produces forecasts, advice, or choices affecting real or virtual settings. The statute adds that such systems use machine and human-based inputs to perceive environments, turn those perceptions into models through automated analysis, and use model inference to formulate options for information or action.
Read closely, the definition separates a system that builds and applies a model from one that follows a fixed script. A payroll program that multiplies hours by a rate is automated. A system that learns from past cases which applications are likely to be complex, then ranks new ones, matches the definition. Many real tools sit between those poles, which is why the label alone settles little.
Canada's test: does it assist or replace judgment?
The Treasury Board of Canada Secretariat publishes a guide on the scope of the Directive on Automated Decision-Making, which governs federal institutions. The guide says the directive is not limited to AI and applies to systems built with rules, regression, machine learning, neural networks, and other technologies. The question is what the system does: if it assists or replaces human judgment, discretion, or critical thought, it is likely in scope, even when the technology looks simple.
The guide's in-scope examples include triaging applications by complexity with machine-defined criteria, generating client scores or classifications, recommending options to a decision-maker, and shortlisting candidates with a résumé-screening tool. Out of scope are pure calculations from published prices and formulas, spam filters, a chatbot that only points clients to government websites, a generative AI tool that summarizes a news article for team members, and systems that apply plain criteria set out in law, such as an age limit the regulation requires.
New York City's three-part test
New York City's Department of Consumer and Worker Protection explains that an automated employment decision tool is a computer-based tool that does three things at once: it uses machine learning, statistical modeling, data analytics, or artificial intelligence; it helps employers or agencies make employment decisions; and it substantially assists or replaces discretionary decision-making. The department describes those techniques as mathematical methods that generate a prediction or classification and work out which inputs matter and how much, in order to improve accuracy.
The wording shows a drafting choice. By listing statistical modeling and data analytics next to AI, the rule reaches tools that vendors may not market as AI at all.
One task, three designs
The table is an illustrative example built from the sources above, using a hypothetical benefit-application triage task. The classifications are a reading of the published definitions, not rulings, and a lawyer or privacy office would decide a real case.
| Design | AI under the US federal definition? | In scope of Canada's directive? |
|---|---|---|
| Reject applicants under 18 because the regulation says so | No, a fixed rule with no model | No, it applies criteria set in law |
| Score applications with weights that staff chose by judgment | Probably not, since no model is inferred from data | Likely yes, the guide treats rules written with judgment as in scope |
| Rank applications with a model trained on past cases | Yes, it infers outputs and recommends | Yes, it assists a decision |
Questions that sort a tool quickly
- Does the output influence a decision about a person, or does it only compute a number from fixed inputs?
- Who set the rules: a person, a law, or a model fitted to data?
- Can a human override or review the result, and is that recorded?
- Which rule applies to the setting, such as a federal department in Canada or a hiring tool in New York City?
This page gives general information about definitions. It is not legal advice, and whether a particular system is covered by a rule is a question for the organization's counsel or the regulator.
FAQ
Is automated intelligence the same as artificial intelligence?
No standard treats them as identical. Automation describes performing a task without step-by-step human action. AI, in the US federal definition, is narrower: a machine-based system whose outputs are predictions, recommendations, or decisions aimed at goals people set.
Does Canada's directive apply to private companies?
The Directive on Automated Decision-Making is a federal government policy for federal institutions. Private-sector systems are outside its reach, although other laws may still apply to them.
Can a simple rules-based system fall under rules for AI?
It can fall under rules for automated decisions. The Canadian scope guide covers rules-based systems that assist or replace judgment, and the New York City test names statistical modeling and data analytics.
Why does the distinction matter in practice?
Because duties attach to the category. Depending on the setting, a system may need an impact assessment, a bias audit, or notice to the people affected.
Related reading
- How AI relates to computer science as a field
- Where deep learning sits inside AI
- Automated systems in accounting work
- AI and everyday society
More in AI Basics.
Last verified: October 2026
Sources
- US Code, Title 15, Section 9401, definitions (National Artificial Intelligence Initiative Act of 2020)
- Treasury Board of Canada Secretariat, "Guide on the Scope of the Directive on Automated Decision-Making": source page
- Treasury Board of Canada Secretariat, Directive on Automated Decision-Making and Algorithmic Impact Assessment pages
- New York City Department of Consumer and Worker Protection, "Automated Employment Decision Tools (AEDT) Frequently Asked Questions": source page