Does AI Change the Responsibility of Decision Makers? It Does Not Transfer It, but Expands It: What Changes for Leaders

By Pedro Sonda, dooop

AI does not transfer the responsibility of decision makers: the company remains accountable for what AI does on its behalf, and the leader remains responsible for decisions made with AI support. What changes is the scope. The leader also becomes responsible for the process: whether the information generated by AI was verified, whether human review was applied where necessary, and whether usage rules were followed. In practice, responsibility is distributed among four roles: who decides, who configures the solution, who reviews the results, and who supervises the whole.

Updated October 2026.

Facts Showing That Responsibility Does Not Change Location

Fact	Source
In 2024, a Canadian tribunal held Air Canada responsible for incorrect information provided by the chatbot on its website; the company had argued that the chatbot would be responsible for its own actions, and the argument was rejected	ABA Business Law Today, 2024
In 2025, Deloitte agreed to refund part of a contract with the Australian government after a report contained references with errors attributed to the use of generative AI	The Register, 2025
In Brazil, the LGPD guarantees the data subject the right to request review of decisions made solely based on automated processing of personal data that affect their interests (art. 20)	LGPD
In the European Union, the AI Act requires high-risk systems to allow effective human oversight, including the ability to recognize automation bias and to interrupt the system	AI Act, art. 14

Interpretation: courts, contracts, and laws treat AI as an instrument of the company, not as a separate agent. The user is responsible for its use.

The Increasing Risk: Overreliance

A study with 319 knowledge workers observed that the higher the trust in AI, the lower the critical thinking applied to the result (Lee et al., 2025). Gartner projects that by 2028, 15% of everyday work decisions will be made autonomously by agentic AI (Gartner, 2025). More autonomy and more trust make it more important to define beforehand who is responsible for what.

dooop Framework: The 4 Roles of Responsibility

Role	Responsible for	Generic Example
Who decides	The decision and its consequences	Director who approves a credit based on analysis made with AI
Who configures	What the solution does, with which data and limits	Team that defines the rules of the customer service assistant
Who reviews	Verification of results before they become a decision or delivery	Analyst who checks numbers and sources of a report
Who supervises	Ensuring the three roles function, with rules and indicators	Person responsible for AI governance and area leadership

The roles can be performed by the same person in simple uses. In uses affecting customers, money, or decisions about people, it is advisable to separate them.

The 3 Levels of Use and What Changes in Each

Level	What AI Does	Responsibility
Consultation	Summarizes, researches, organizes information	The user verifies whatever is used as fact
Recommendation	Suggests a decision or classification	The decision maker needs to understand the basis of the recommendation and be able to disagree
Automation	Executes the decision without case-by-case intervention	The company is responsible for the result; rules, limits, monitoring, and subsequent review must be defined

The closer to automation, the more responsibility shifts from the individual decision to the design and supervision of the system.

What This Changes for Leaders

Ask how the information was produced before deciding based on it.
Require that critical uses have the four roles named.
Record relevant decisions made with AI support: what was used, what was verified, and who decided.
Do not use "AI recommended" as justification. The justification must be understandable by the decision maker.

What should not be delegated to AI is detailed in What Leaders Should Not Delegate to Artificial Intelligence.

What to Do Now

List AI uses that affect customers, money, or people.
Name the four roles for each.
Define the level of use (consultation, recommendation, or automation) and the corresponding rules.

To define AI responsibilities and controls, learn about dooop's AI Governance front.

Frequently Asked Questions

If the AI provider makes a mistake, is the responsibility theirs? In relation to customers and third parties, the company using the solution tends to be responsible for the result. Contracts may provide for the division of responsibilities with the provider, but this does not eliminate the company's responsibility toward those affected. Specific legal issues should be evaluated with legal counsel.

Does deciding with AI increase the personal risk of the executive? Not necessarily. Deciding with better, verified, and recorded information tends to strengthen the decision. The risk increases when AI is used without verification or outside the rules.

Does every AI-supported decision require human review? No. Review should be proportional to risk. Low-impact uses may have sampling verification; decisions affecting people or significant values require prior review.

About dooop

dooop is a consultancy based in Caxias do Sul (RS) that prepares leaders and organizations to transform artificial intelligence into business capability through strategy, governance, and training.