top of page
Writer's pictureDigital Team

A framework for an AI strategy


Framework structure

The 2019 OECD framework for an AI strategy (The Framework) provides a helpful structure for the development of a national AI strategy.


The Framework suggests breaking the approach into the 'baseline' - which is an assessment of the organisations current strategic situation and challenges that AI might help address; 'objectives' - what the organisation wants to achieve using AI and the principles that will underpin the actions it takes to achieve them; and 'approaches' - which are the concrete actions that will be undertaken to achieve these objectives.


A good strategy will require active monitoring for a clear overview. The elements should also be developed concurrently and iterated as the context evolves


Baseline components


Determine current strengths and weaknesses by mapping:

  • Internal AI capabilities.

  • Government and external data assets.

  • Existing government AI and data science projects.

Assess the strategic context:

  • Public and workforce attitudes to government and AI .

  • Current legal and regulatory framework.

  • Existing government and international commitments, institutions.

  • Academic and private sector expertise that might be drawn upon.

Identify specific operational problems that AI has the potential to help solve:

  • Adopt a multi-disciplinary and diverse approach to decide whether AI is the best solution to a policy problem, and if so, how it should be designed and implemented.

  • Put in place processes to understand user needs.

  • Create mechanisms to match resources to priority problems.

  • Define the specific decision AI will make or support.

  • Consider who will be impacted by this decision and associated risks if it fails.

  • Explore how the service will need to be redesigned to leverage the impact of AI.


Objectives


Decide what goals the AI should help government achieve:

  • Articulate how AI will generate public value and specify missions to which AI can be part of the solution.

  • Engage stakeholders in goal definition.

  • Make space for experimentation and learning.


Define and communicate to stakeholders the principles that will shape how AI is used in government:

  • Fairness and unbiasedness.

  • Transparency and accountability.

  • Privacy and individual autonomy.


Approaches


Ensure government access to AI capability and capacity:

  • Construct talent pipelines, and develop recruitment and retention plans for internal technical expertise.

  • Harness external expertise through partnerships and collaboration.

  • Design effective public sector AI procurement processes.

  • Build a cadre of civil servants who understand the legal, ethical, technical and managerial issues around AI.

  • Establish funding schemes and secure the availability of resources in the government’s fiscal plans.

Secure ethical access to, and use of, quality data and infrastructure:

  • Determine what data are needed to address the problems.

  • Decide how to obtain input data of sufficient quality and that are sufficiently representative of the target population to make accurate predictions with minimal bias.

  • Develop a data strategy that complies with data protection law and best practice and is consistent with AI principles.

  • Ensure important data infrastructure are in place (e.g., hybrid cloud services).


Put in place legal, ethical and technical frameworks to operationalise the principles:

  • Monitor compliance with principles during implementation to track progress, and identify and respond to emerging issues.

  • Put in place safeguards against bias and unfairness.

  • Clarify the appropriate role for humans in the decision-making process.


Develop open and transparent accountability structures.


Prepare for future shifts:

  • Ensure openness and flexibility in future plans and contracts.

  • Follow OPSI Anticipatory Innovation Governance concepts


Conclusion


Countries developing an AI strategy could consider using the OECD framework for an AI strategy as a helpful organising structure. The three parts (baseline, objectives, approaches) should all be considered in parallel and with the expectation that the output (strategy) will be a highly dynamic guideline for AI across government and beyond.


AI framework

References:



OECD (2019c), “Using digital technologies to improve the design and enforcement of

public policies”, OECD Digital Economy Papers, No. 274, OECD Publishing, Paris,

https://dx.doi.org/10.1787/99b9ba70-en.


European Commission (2018a), Artificial Intelligence: A European Perspective,

European Commission, Brussels, https://ec.europa.eu/jrc/en/publication/eur-scientific-

and-technical-research-reports/artificial-intelligence-european-perspective



bottom of page