Laurent -

Government publishes artificial intelligence procurement guidance

A document has been published outlining the challenges for public sector buyers as well as best practices on procuring AI technology

The UK government has published a document outlining the main challenges around artificial intelligence (AI) procurement for public sector buyers as well as best practices.

Developed by the Office for Artificial Intelligence (OAI) in collaboration with the World Economic Forum (WEF) Centre for the Fourth Industrial Revolution, Government Digital Service (GDS), Government Commercial Function and Crown Commercial Service, the document seeks to enable public bodies to buy AI systems in a more confident and responsible manner.

The guidance for procurement, which refers mostly to the use of machine learning and was created with the input of a range of stakeholders from industry, academia and government departments, was initiated through the WEF’s Unlocking Public Sector AI project. It follows a previous guide to using AI in the public sector by the OAI and the Government Digital Service, released in January 2020.

As part of the initiative to help managers in government in terms of decisions around buying AI systems, the OAI brought stakeholders together to develop AI Procurement in a Box, a toolkit to help procurement professionals develop their approaches to AI procurement in the UK and abroad.

The document provides an overviews of what themes should be considered when assessing the possibility of buying an AI system and is intended for use alongside other previously published policy and advice, in areas such as data ethics and open data standards.

Including procurement in a strategy for AI adoption is one of the 10 items of advice listed in the document. Here, buyers are told they should be using procurement strategically to encourage AI adoption across government, take advantage of economies of scale in the public sector and set up networks to learn from work done elsewhere. Setting up diverse teams to ensure understanding of the interdependent disciplines that AI technologies incorporate is also part of the advice.

The guidance noted that buyers should conduct a data assessment before starting a procurement process for an AI system, and consider asking questions such as whether relevant data will be available for the project. Benefits and risks of AI deployments should also be considered, as well as asking whether AI is the right solution for the problem at hand.

Effective market engagement is another piece of advice provided in the guidance. According to the document, procurement professionals should be careful with actions that could introduce hurdles for certain under-represented groups, including small and medium-sized enterprises (SMEs), as well as voluntary, community and social enterprises suppliers from competing. Technical and ethical shortcomings of AI deployment should also be considered, the document says.

Carefully writing the project requirements to help suppliers understand what is required is another part of the guidance, which suggests exploring different routes to market to acquire AI systems – for instance, Innovation Partnerships, the GovTech catalyst and the Crown Commercial Service’s AI Dynamic Purchasing System.

Developing a plan for governance and information assurance is another part of the advice provided to AI buyers in the pubic sector, as well as avoiding risks such as supplier lock-in. The need to consider lifespan testing for AI deployments, rather than one-time solutions, is also stressed in the guidance.

In addition, the document raises specific considerations to be addressed during processes of acquisition of AI systems. These include preparation and planning, selection, evaluation and awards, as well as what to keep in mind at the stage of contract implementation and ongoing supplier management, including knowledge transfer and training, and end-of-life processes.

Read more about artificial intelligence

Read more on IT for government and public sector

Data Center
Data Management