Docs

Appearance

"Enable end-user responsibility" ​

"Enable end-user responsibility" is dimension 4 of the IMDA Model AI Governance Framework for Agentic AI (MGF for Agentic AI), in section 2.4. In this dimension, the MGF for Agentic AI recommends information, education, and training for end users.

Types of users ​

Section 2.4.1 of the MGF for Agentic AI recommends that organizations "cater to different users with different information needs".

The MGF for Agentic AI is voluntary guidance from IMDA, not law.

Section 2 of the MGF for Agentic AI tells organizations that "trustworthy deployment of agents does not rest solely on developers, but also on end-users". In section 2.4, end users "are the ones who use and rely on agents, and human accountability also extends to these users". The MGF for Agentic AI recommends that organizations "provide sufficient information to end users to promote trust and enable responsible use".

Section 2.4 gives two items that "Organisations should consider":

  • Transparency — "Users should be informed of the agents' capabilities (e.g. scope of agent's access to user's data, actions the agent can take) and the contact points whom users can escalate to if the agent malfunctions."
  • Education — "Users should be educated on proper use and oversight of agents", and also on "the potential loss of trade craft". For tradecraft, see Tradecraft and business continuity.

Section 2.4.1 gives two "main archetypes" of end users:

Type of end userExamples in the MGF for Agentic AIRecommended in section 2.4.1
"those who interact with agents" (section 2.4.2)Agents that "act on behalf of the organisation", for example "customer service or sales agents". These agents "tend to be external facing, although they can also be deployed within the organisation", for example "a human resource agent that interacts with other users in the organisation"."Focus on transparency". See Information for users.
"those who integrate agents into their work processes or oversee them" (sections 2.4.1 and 2.4.3)These users "typically utilise agents as part of their internal workflows", for example "coding assistants, automation of enterprise processes". "The agent acts for and on behalf of the user." These agents are "mostly internal-facing".The same information, and also education and training. See User training.

The external and internal examples do not limit the two types.

For the continued accountability of organizations and persons, and the information for users in §2.2.1, see Human accountability and Agent value chain.

Information for users ​

Section 2.4.2 of the MGF for Agentic AI gives the information for users who interact with agents. For these users, the MGF for Agentic AI tells organizations to "focus on transparency". Organizations "should share pertinent information to foster trust and facilitate proper usage of agents". The MGF for Agentic AI gives five items of this information, and it uses "can include" for these items. Thus the five items are examples, not a complete set.

The Executive Summary of the MGF for Agentic AI gives a "baseline" of three items. It recommends that users get information about the "range of actions" of the agent, its access to data, and "the user’s own responsibilities".

Section 2.4.2 gives these five items:

"User's responsibilities"

Clearly define the user's responsibilities, such as asking the user to double-check all information provided by the agent. Where appropriate, allow users to set their own approval thresholds and boundaries for agent actions based on their individual risk tolerance, beyond organisation-defined limits.

"Interaction"

Declare upfront in the user interface that the users are interacting with agents at the point of interaction, rather than writing it only in separate documentation.

"Agents' range of actions and decisions"

Inform the users on the range of actions and decisions that the agent is authorised to perform and make.

"Data"

Be clear on how user data is collected, stored, and used by the agents, in accordance with the organisation's data privacy policies. Where necessary, obtain explicit consent from users before collecting or using their data for the agents.

"Human accountability and escalation"

Provide users with the respective human contact points who are responsible for the agents, whom the users can alert if the agents malfunction or if they are dissatisfied with a decision.

The MGF for Agentic AI gives a location only for the "Interaction" item: "in the user interface" and "at the point of interaction". It gives no location for the other four items.

An example from Workday ​

The MGF for Agentic AI gives an example from Workday about its "finance and HR agents". This example shows how one organization applies transparency to end users. It is an example, not a rule for all organizations. In this example:

  • The user interface "clearly provides notice to users that the tool is powered by AI and is intended to support a select range of actions whenever a user interacts with an AI agent e.g. in Slack or Microsoft Teams".
  • The "factsheets provided to users" give the tasks that an agent can do, for example "reviewing candidates against job requirements" and "summarising talent profiles". These "factsheets" also give the tasks that the agent cannot do, for example "making decisions related to hiring promotion or succession" or "initiating changes to worker records". This shows "that users and HR remain in control".
  • In sensitive workflows, the agents "accompany each recommendation with an explanation of their reasoning, including the data considered, key factors that influenced the recommendation, and any noted uncertainties or risks".

User training ​

Section 2.4.3 of the MGF for Agentic AI recommends that organizations give education and training to users who "integrate agents into their work processes". These users also get the information for users.

In section 2.4.3, the "key aspects" of the education and training "include" these items:

  • "Foundational knowledge on agents":
    • The "relevant use cases", to help users integrate agents into their "day-to-day work". Also "the scenarios under which the use of agents should be restricted (e.g. do not use an agent for confidential data)".
    • How to give instructions to the agents, for example "general best practices in prompting".
    • The "Agents' range of actions", to help users know the "capabilities and potential impact" of the agents.
  • "Effective oversight of agents":
    • "Common agent failure modes, such as hallucinations, getting stuck in loops after errors", to help users "identify and flag out issues".
    • "Ongoing support, such as refreshers to update on latest features and common mistakes".
    • "Feedback loops": "when users override agent actions or identify wrong actions, enable them to report this such that it can be used to improve the agent processes if needed".
  • "Potential impact on tradecraft and business continuity": see Tradecraft and business continuity.

For the training and domain expertise for human oversight, see Human oversight.

Tradecraft and business continuity ​

Section 2.4.3 of the MGF for Agentic AI gives these risks to the tradecraft of users and business continuity:

  • "As agents take over entry-level tasks, which typically serve as training for new staff, this could lead to skill degradation, in which users lose basic operational knowledge."
  • "This can create business continuity risks in which users may no longer know how to perform critical processes manually when agents malfunction or become unavailable."

For these risks, the MGF for Agentic AI recommends that organizations "identify core capabilities of each job and provide sufficient training and work exposure so that users retain foundational skills".

Section 2.4.3 gives tradecraft as an item that is different from the "Foundational knowledge on agents" and the "Effective oversight of agents".

For the limits and controls for agents that do not operate correctly, see Limits on agents and Deployment and monitoring.

Use the framework in Modulos ​

In Modulos, the AI System Framework MFF-17 and the Organization Framework OFF-17 contain Requirements for the MGF for Agentic AI. For the Requirements for sections 2.4.2 and 2.4.3, and for the steps that Modulos recommends, see Use the MGF for Agentic AI in Modulos.

For the MGF for Agentic AI, see also:

Sources ​

  • MGF for Agentic AI — IMDA, Model AI Governance Framework for Agentic AI, Version 1.5, "Published 20 May 2026 (Updated 5 June 2026)". Provisions: the Executive Summary, section 2, section 2.2.1 (End users), sections 2.4 to 2.4.3 with the Workday example, and sections 2.1.2 and 2.3.3. IMDA PDF.

Disclaimer

This page gives general information. It is not legal advice.