Making the case for public participation in AI procurement

The principles for Public Participation in AI Procurement we will be co-launching alongside the Open Government Partnership Summit in a few weeks time focus on the what and the how of involving affected communities in decisions at all stages of buying and managing AI systems in the public sector. But ahead of that, we need to answer the question why? Why is is particularly important to involve publics in AI procurement, when, to date, procurement has often not been a space where public engagement is widespread.

Preparing for our workshop at the Open Gov Hub today, I’ve jotted down the following reasons.

AI is an Emerging Technology

AI is not like tables and chairs, office supplies or the regular kinds of services public authorities might buy in so far as both the benefits, and the risks, are often highly uncertain.

AI is often put forward as holding significant promise to save funds and transform services, yet in these early days of adoption evidence is limited as to how much of that benefit will be delivered, and what harms or opportunity-costs might be encountered along the way.

In this context, it’s vital to have different voices at the table - and involving the public can help authorities to understand the opportunities and costs or risks on an AI procurement for different groups, and to be better equipped to weigh up trade-offs in an environment of uncertainty.

Over time, certain forms of AI might become more ‘normal technologies’ where there is clearer societal understanding of their costs and benefits: and it’s possible to see that public engagement in these contexts might become simpler and more routine - restricted perhaps to moments of oversight. But even established technologies introduced into new communities or into new contexts can create uncertainties that demand public engagement to explore.

The other important aspect of emerging technologies is that they are still being shaped, and so we face a particular moment of opportunity right now to shape how AI will impact our societies by bringing public voices to the table when the buying power of public authorities is deployed.

AI is Infrastructure

In Ismael Pena-Lopez’ ‘Stakeholder Engagement in Public Procurement for Artificial Intelligence: a Mission-Oriented Playbook, commissioned by ParticipationAI, Ismael makes the argument that we should treat AI as public Infrastructure. That is, AI is not a tool that just drops into existing workflows and processes - but generally any significant enough investment in AI reshapes working practices, and becomes an infrastructure on which future services and decisions are built.

Choices around infrastructure are significant choices with long-lasting consequences. Decisions about the planning of a new neighbourhood, or building a new road or transit link, often have consequences for decades to come - and generally involve a good deal of public engagement to get the decision right.

Choices about AI infrastructure, whilst perhaps rarely involving the capital outlay or sunk costs of a new highway, can shape an area, or the lives of a particular group, in ways that those affected have a right to input upon. And unless planned for from the outset, vendor lock-in can leave technical infrastructure as embedded, brittle and difficult to change in future as infrastructure in the built environment.

For example, the introduction of AI-based case notes and case management into social work reshapes the daily practices of social workers, but also has impacts on consequential decisions affecting the life-histories of children and families. And unless the potential to modify or exit to alternative systems has been designed for from the start, AI tools delivered by large firms can be far harder to adapt during operation in response to worker, client and management feedback than systems that were developed closer to the ground.

In such cases, where there is evidence that the benefit of the infrastructure will be high, but that iterative development will be difficult, up front public engagement to get the design right becomes vital.

AI can displace Agency

When AI is used for automated decision making, then a certain degree of agency that was previously held by, hopefully accountable, human beings is transferred to the developers or AI systems, and the models they have created. If we don’t want to see a fall in the effective accountability of authority decision-making, then we need new mechanisms of public engagement and oversight through the buying process.

If we go back to the example of the AI-based case management in social work. When case management is relatively analogue, and some issue arises that needs a change in practice (e.g. feedback from clients or workers that earlier intervention on some issue would be beneficial), then change can happen through a dynamic negotiation of clients, frontline staff and management. When workflows are built into digital and AI systems, the route to create change can narrow to management through the technology contract.

The 2025 UNDP Human Development Report makes the case for AI as a complementary technology, that should expand, rather than constrain, human agency. But to secure that, requires careful oversight and engagement in choosing and configuring it - and that needs a diversity of voices in the room during procurement and management.

Public involvement should be a proportionate part of all procurement

The reasons above show why public engagement in AI procurement is particularly important at this moment, and the principles below outline some of the specific considerations when engaging publics around AI, such as the need for careful capacity building, and independent expert inputs. However, we should not push the AI exceptionalism to far: proportionate public engagement can be valuable across all forms of procurement - and many of the ideas explored below could equally apply to infrastructure, product and service procurement across the public sector.

In  Stakeholder Engagement in Public Procurement for Artificial Intelligence: a Mission-Oriented Playbook, Ismael also makes the case that stakeholder engagement around AI procurement plays an important role in building public trust, and managing the complexity of modern technical systems through bringing diverse voices into the mix. Recognising also that often public engagement may be framed broader than a particular buying decisions (e.g. input to the pre-procurement phase could come from public dialogue about wider aspects of service reform), participation can play a powerful role in aligning technical, organisational and community change.

What reasons resonate for you?

Do the reasons above resonate for you? Or are there other arguments you think need to be in the mix?

Drop me a line or join discussion on LinkedIn here.