Standard guide that is being developed by the National Office of Information Technologies to guide the design and sizing of projects incorporating Artificial Intelligence Agents in public administration. It is only aimed at national public administrations seeking to adopt this technology in a planned and responsible manner. Its objective is to disseminate best practices, reduce risks, and ensure secure, effective implementation aligned with State priorities and regulations.
Initiative overview
The initiative consists of a standard guideline developed by the National Office of Information Technologies (ONTI), as the governing body, to guide the design and sizing of projects that incorporate Artificial Intelligence Agents in public administration. It arises from the lack of common criteria for the development of these systems and the risk of isolated implementations with proprietary standards, design flaws, or deficiencies in security and governance. It also responds to the need to organise the growing adoption of generative AI-based tools within the National Public Administration.
The guideline aims to establish a common framework that promotes open standards and interoperability, incorporates security, governance, and human-in-the-loop oversight criteria, and facilitates informed decisions regarding the levels of autonomy of the agents. It also seeks to optimise the use of public resources, strengthening the efficiency and effectiveness of technology projects through best practices in architecture, risk assessment, and results measurement.
While not mandatory, its use is recommended as a best practice guide for the design and sizing of initiatives involving AI Agents. It will be integrated into formal, non-binding technical evaluation processes and updated based on technological advancements. Furthermore, it is complemented by dissemination and training activities targeting technical, legal, and procurement areas, and can serve as a reference for other levels of government, given the guiding role currently played by ONTI's technological standards.
Other relevant details
Results, outcomes and impacts: Since the guideline has not yet been formally adopted as a consolidated reference standard, its impacts are primarily anticipated rather than immediately observable. The initiative is designed to strengthen technical and institutional capacities in the design and sizing of AI Agents within the National Public Administration, and promote common criteria based on open and interoperable standards. Expected results include better integration of human-in-the-loop oversight mechanisms, greater clarity in defining responsibilities and levels of autonomy, improved traceability in agent-assisted processes, and greater institutional capacity to adopt AI-based solutions securely, efficiently, and in alignment with government priorities.
Challenges and lessons learned: A key difficulty has been balancing the technical precision required for effective AI Agent design with the clarity needed for a practical and accessible reference guide for public organisations. Another challenge is organising and systematising the various approaches used to classify and describe this rapidly evolving technology, ensuring that components, levels of autonomy, and design criteria can be clearly understood. This process highlights the importance of building a coherent framework with consistent terminology and well-defined categories to support responsible and coordinated adoption in public management. The main lesson learned is that a common framework should not only establish principles but also offer an organised conceptual architecture that facilitates technical and strategic decision-making. Consistent terminology, well-defined categories and explicit design criteria are essential to promoting responsible, coordinated, and sustainable implementations in the public sector
