automated decision-making (ADM)

basicwidespread

A decision made about you by a computer system based on your data, without meaningful human intervention.

In plain terms

Imagine that in an office there is no clerk sitting behind the window, but an automaton: you feed in your data, and the answer comes out saying approved or rejected. No one looked you in the eye, no one weighed the options. This is automated decision-making: the machine not only prepares the decision but also makes it.

It is used because it is fast, cheap, and works uniformly at scale. A bank can process thousands of loan applications in minutes, while a large company can screen tens of thousands of resumes. The machine does not get tired and, in theory, applies the same rule to everyone.

As a legal concept, it is important because in the European Union you receive special protection if such a decision significantly affects your life—for example, concerning a loan, job, insurance, or benefit. Such purely automated decisions are generally prohibited and permitted only in a few exceptional cases. Even then, you can typically request human review, state your position, and challenge the decision.

Not every automation falls under this. A movie recommendation or spam filter is also a machine decision, but it has no serious stakes. The regulation focuses on cases where the decision has legal or similarly significant consequences for you.

How it works

The system collects data about you: what you provided (income, age, education), and what it obtains from elsewhere (payment history, shopping habits, behavior in an application).
The system creates a score or classification from the data. This can be provided by human-written, fixed rules, or a machine learning model trained on past cases. If the system thus evaluates your personal characteristics or predicts your behavior, it is called profiling.
The score is measured against a threshold: above it, approval; below it, rejection, or perhaps a higher price, lower ranking, or account lockout.
The decision takes effect without human approval. If a person merely mechanically nods along with the machine's suggestion, it may still legally count as an automated decision.
In the EU, the decision-making body must inform you that a machine made the decision and on what logic, and as a general rule must provide the opportunity to request human review.

An everyday example

At 10 p.m., you apply for consumer credit on your phone for a new washing machine, and the rejection arrives in half a minute. No human would have reviewed your application that quickly: a scoring system made the decision. You have the right to ask on what basis and request that a case officer also examine it.

Who is affected

bank customer, investorCredit assessment, credit limits, and card blocking due to suspicious transactions are now largely machine decisions. → Money

jobseeker, new graduateAt many large companies, software filters resumes, so your application may never be seen by a human. → Work & HR

entrepreneur, executiveIf your company entrusts machines with decisions affecting clients or staff, you have an obligation to inform and review; failure may result in a fine.

lawyer, citizenAuthorities also handle more cases automatically, so you must know when you can request human review. → Law

A common misconception

MythThe machine is objective, so its decision is fairer than a human's.

In realityThe system works from past data and rules given to it by humans, so it can learn and massively repeat old biases. Consistency is not the same as fairness.

Limits and risks

The ban is not total: purely automated decision-making is permitted if necessary for the conclusion or performance of a contract, if expressly allowed by law, or if you have explicitly consented. In the case of a contract and consent, you may request human review. If a law permits automated decision-making, that law determines what protection you receive.
Often you do not even know a machine made a decision about you, and the explanation received is often generic. Companies treat the exact calculation method as a trade secret, although the Court of Justice of the European Union ruled in 2025 that a trade secret alone does not exempt one from providing an understandable explanation.
Flawed or outdated data leads to flawed decisions, so it is worth periodically checking what is recorded about you.
Formal human approval means little if the caseworker, short of time or authority, always accepts the machine's suggestion.

Similar, but not the same

Profiling – Profiling only evaluates and categorizes you based on your data, whereas automated decision-making also makes consequential decisions, sometimes without profiling, based on simple rules.
Decision support system – A decision-support system only makes a recommendation, and the final decision is made by a human who genuinely exercises judgment.
AI-agent – An agent independently executes a task in multiple steps, whereas automated decision-making is a legal concept referring to a decision about you that affects you, regardless of the technology used.

Related concepts

Broader

data protection, algorithmic governance, regulation of artificial intelligence

Narrower

credit scoring, Automated resume screening, algorithmic work organization

On the edge

Profiling, human oversight, algorithmic bias, Explainability, High-risk AI system

A brief history

The 1995 EU Data Protection Directive already included protection against decisions made solely by automated means. The current rule is codified in Article 22 of the General Data Protection Regulation (GDPR), applicable since 2018. In 2023, the Court of Justice of the European Union ruled in the SCHUFA case that the calculation of a credit score itself may qualify as an automated decision if the creditor's decision depends to a decisive extent on it. In 2025, in the Dun & Bradstreet Austria case, the Court also established that the data subject is entitled to a comprehensible explanation regarding the procedure and principles on which the decision concerning them was made.

Where things stand

This is common practice in finance, recruitment, insurance, online platforms, and public administration. Alongside the GDPR, the EU Artificial Intelligence Regulation also sets requirements for high-risk systems and grants the right to request an explanation for decisions made by such systems that significantly affect the individual. The application of these rules has been postponed from the original August 2026 deadline to later dates, with most affected systems now facing a December 2027 deadline. The focus of legal disputes has shifted to what constitutes meaningful human intervention and how detailed the explanation provided to the data subject must be.

Status: 2026 November

In the news

Appears in 4 Agent Age stories; the curve shows weekly occurrence in the news magazine.

UK GDPR still applies to AI after the Data (Use and Access) Act, NeuralTrust saysOct 22, 2026 · Jog
EDPB publishes €824.99 million GDPR fine against Uber; Noyb cites €2 billion in finesOct 12, 2026 · Jog
US state and federal AI rules collide: Connecticut, Colorado, FTCAug 17, 2026 · Jog
Colorado proposes automated decision-making rules; ABA calls for federal AI regulationAug 14, 2026 · Money

Also known as

automated decision-making · automated individual decision-making · automated decision · Algorithmic decision-making · machine decision-making · automated decision-making · ADM · algorithmic decision-making

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.