The case for a local government AI taskforce

The UK government has made known its hope for an AI-driven improvement of public services. The AI Opportunities Action Plan and A blueprint for modern digital government signal just how high the expectations are for data-intensive technologies to fix the public sector and address the societal and financial challenges it faces. But deploying AI comes with its own issues.

The National Audit Office’s reports on digital transformation and the government’s relationship with technology suppliers highlight the existence of power imbalances between public institutions and private tech providers, leaving the former at a disadvantage. This is an important issue also given the amount of money that the public sector spends on procurement of goods and services.

Within local government in England, continuous budget cuts and the proposed devolution of power are unfolding alongside debates about how government’s public sector initiatives towards digitalisation and AI implementation may or may not be aligned with local government needs.

For instance, responding to the government’s vision for a digital transformation of public services, the Local Government Association (LGA) has argued that ‘councils have a unique position within the public sector technological eco-system’. Local administrations can support central government to achieve the objectives of its digital reform of public services and make them more efficient. A successful digitisation of government, however – the LGA has noted – also means recognising that local government needs tailored approaches instead of a one-size-fits-all solution. These approaches must consider the specific role of local vis-à-vis central government, and the necessities of and differences between different local authorities.

The way in which the public sector buys AI must ensure that its use and the outcomes it leads to address the needs of people and communities and are fair, transparent, legitimate, safe and trustworthy. But many of the decisions to be made around AI implementation will also need balancing against financial and broader economic policies, which may challenge those objectives.

If its adoption continues in domains that range from administrative tasks to service delivery, AI will increasingly play a role in the running of public services – even mediating the relationships between people and the public institutions that represent them. Here, the stakes will go beyond ensuring technological tools work and deliver value for money: inappropriate or uncritical AI adoption could seriously impact people’s lives.

The Ada Lovelace Institute’s year-long, collaborative research finds that local government procurement of AI technologies could be the place to start to strengthen responsible public sector adoption. More specifically, getting AI procurement right at the local level could positively impact local communities, while supporting central government’s objectives around better and more efficient public services.

The state of AI procurement in local government

Our research finds that local government is not sufficiently supported in this process.

Clarifying the purpose and role of AI in relation to the general mandate of the public sector to act in the public interest is a crucial issue to address. The analysis we conducted on legislation and procurement guidance documents, however, found that terms such as public interest, transparency, fairness and social value do not have clear definitions and there are no ways of operationalising them in real life decision-making.

Alongside these uncertainties, local government procurers navigate practical challenges that relate to AI governance and technological and infrastructural unpreparedness. The technology and procurement experts we interviewed across industry and local and central government detailed the imbalance of skills and knowledge between the private and public sectors, which sometimes leaves the latter unable to critically engage with companies’ claims about AI products. Additionally, market monopolies affect the relationships between buyers and sellers in ways that may not be suitable for procuring and deploying responsible AI in the local public sector.

During our research, we found that institutional structures meant to support local and central government procurement can sometimes complicate the route to buying AI. For instance, the Crown Commercial Service (CCS) coordinates and establishes standards for central government procurement. There is no single office with the same kind of coordinating and oversight mandate for local government. Different actors share the responsibility of setting and checking standards, leaving procurers uncertain if and how to apply CCS frameworks to their processes. 

In parallel with these issues, our research also covered examples of excellent practice, with teams working on collective AI purchases, sharing knowledge and producing toolkits to support contract management or ethical decision-making. Some of these efforts are regionally coordinated, but for the most part they are driven by local circumstances and polices. The knowledge base they develop is not always openly shared, which leads to the duplication of efforts, as well as unequal practices, use and outcomes across regions.

Building on local experience to ensure procurement of legitimate AI

Improving procurement to ensure AI serves people and communities means establishing the role of AI in specific services and contexts. It means empowering local administrations to choose what to focus on when they procure – where they may prioritise serving a certain demographic, expediting replies to requests, managing resources, etc. – according to local needs. It also requires central government to intervene in market dynamics – with measures that counter the dominance of a small number of companies and vendor lock-in effects – so local government can choose how and from whom it sources AI technologies. Wider regulatory support is also needed to help local government to address gaps in AI governance through the creation of practicable guidance. Without this, local officials will struggle to make confident choices about AI that work for the service and their communities.

Enabling good procurement processes entails a delicate balancing act between centralising certain functions and responsibilities and avoiding top-down decision-making, which imposes standardised solutions and leaves local government out of the conversation. It requires dedicated resource and support from central government, with input from industry where appropriate.

The results of our research led us to call for a national taskforce to reform local government procurement of AI, comprising of representatives from both local and central government.

Recently we welcomed the Department for Science Innovation and Technology’s commitment to engage more deeply with local government, which resonates with our call, especially in light of initiatives such as the AI Opportunities Action Plan, AI Management Essentials toolkit and the AI Playbook to support the use of AI in public services. However, a roundtable discussion we convened with those working on and supporting local government procurement showed where these recent initiatives might fall short.

For example, the government has stepped up the focus on digital and AI-related skills in e-learning modules for civil servants and Parliament and set up the Parliament’s digital and data skills centre of excellence. But these programmes exclude local government offices and public authorities, where training is also needed. The omission results in running similar skill acquisition courses at both national and local levels, duplicating efforts and diverting training resources from where they are needed.

While they are positive steps towards improving procurement, the government initiatives mentioned above seem to assume that the public sector is homogeneous and end up missing the nuances and specific considerations necessary for successful digital transformation and AI deployment in local government. This could lead not only to the duplication of efforts but also to the adoption of policies that are unlikely to work, along with the waste of already stretched budgets.

Local and central government should capitalise on opportunities to work more closely together. Local government has practical knowledge of the needs of their communities and how to deliver the necessary services. But when it comes to wider policymaking, it faces various challenges and a lack of power to drive change where industry practices and central government buy-in are key for success.

Only if the knowledge cultivated by local authorities is represented at the national level, in a collaborative partnership between central and local government institutions, will AI procurement at the local level improve and deliver concrete benefits for people. And good procurement policy decisions can help address and curb the financial costs and burdens accruing from current digital and AI investment.

Conditions for success

To explore the possible shape of a taskforce, we convened a roundtable with regional and local government experts, representatives from academia and civil society. We also invited feedback from regulators, industry and central government representatives.

The roundtable confirmed the ongoing need for a local-central government partnership, although, to be successful, a taskforce will need to be iterative and remain relevant as AI technologies and policies continue to transform. A taskforce could start by focusing on the following outcomes:

• Aligning the goals of central and local government, factoring in the contextual needs of local authorities and making room to analyse and address them directly.

• Ensuring technology deployment programmes gain and are worthy of the trust of the communities affected by them, with local government confidently holding everyone involved in supplying technological products and delivering services accountable for the process.

• Enabling councils to share ideas among themselves, and work and procure together.

• Supporting in concrete ways the application of clear standards across the procurement process.

• Establishing a decision-making model that is attentive to market concentration and its effects on AI procurement (e.g. vendor lock-in dynamics).

Achieving these outcomes means redistributing power from industry to central government, from central government to local government, and from local government to communities. This is a stepwise approach, in which central government must:

• Empower procuring entities with practical tools and resources, and the knowledge required to relate to industry, through targeted training and by providing independent expertise for procurers to seek advice from.

• Hold all third parties working to deliver services to local government accountable to the same public sector standards and democratic values as local government itself.

Conclusion

There is considerable potential in strengthening the procurement of AI at a local level to set the path for implementing technology that supports services and benefits people.

Our research has led us to the conclusion that a more joined-up approach with dedicated resources to support local administrations to procure and use AI well is necessary. The gaps we see in recent announcements on digital and AI transformation strengthen our earlier call for a taskforce to improve the process of procuring AI for local public services.

A taskforce devoted to changing procurement for the better will provide an outcomes-based approach to implementing the necessary changes to reshape the market for AI products for government, support skill development and enable local procurers to do their job well.

We see the opportunities that such a taskforce can bring, including to unify the goals of local and central government in ways that are not duplicative or directive but constructive and collaborative.

We acknowledge and thank the experts who attended the roundtable for their insights which were invaluable to shaping whether and how our recommendation for a taskforce can be developed.