by Jasmine Kendall, Isak Nti Asare, Eleanor Shearer, Richard Stirling and Tom Westgarth

National AI strategies are a useful way for a government to signal that it is taking the opportunities and challenges of AI seriously. The best strategies will also be more than just a signal – they will be able to effectively coordinate and prioritise policy responses that put governments in the best position to make use of AI. 

We have previously noted that if governments are to capitalise on the AI opportunity, they need to have well thought out structures and processes in place to do so. However, despite multiple studies and reports, no one – to our knowledge – has addressed the question or what makes a good AI strategy? “Good” is a subjective term, but here we take it to mean the kind of AI strategy that will actually lead to change. 

– IAIDL has not only been an advocate for AI strategies but has been a part of crafting AI strategies at the national and municipal level. But not all AI strategies are made equal.  As more countries and regions wake up to the realisation that AI could transform their economies, public services, and workforces, and themselves begin to develop national strategies the need to understand what makes a good strategy will increase. What follows are our initial thoughts on some of the key factors that make-up a good AI strategy.    

1. Clear and realistic purpose

First and foremost, a good AI strategy needs a clear vision about what needs to change. Without a clear purpose, a strategy cannot act as an effective roadmap for policy. In setting targets for change, governments need to be realistic. There is no point in a country creating a roadmap to be a world leader in providing an AI service that they quite simply do not have the capital, infrastructure and demand available to deliver on. 

This is where taking a sectoral-focused approach can be particularly valuable. The best AI strategies take into account the environment that a nation finds itself in, and can tailor their approach  in order to fit these conditions.

For example, the Mauritius AI Strategy lists the ‘prioritization of sectors and identification of national projects’ as one of its main focus areas. The strategy highlights a number of different sectors in which AI could support economic growth in the country. The Ocean Economy (which comprises 10% of Mauritius’ GDP) and textile manufacturing (which in 2017 comprised 8.7% of Mauritius’ GDP) are among the sectors listed in the strategy as possible areas of AI application.

Finally, clarity of purpose will also involve an element of prioritisation. In order to prevent a strategy becoming a wish-list of unattainable goals, governments will often have to choose what not to do as much as what to do. Such trade-offs should be made consciously and will usually be informed by the specific strengths of a particular economy. 

2. Clear timeline

It’s relatively easy for a government to make ambitious claims about how it will harness the benefits of artificial intelligence in the future. Staying true to these claims and faithfully implementing a strategy is much harder. 

In order to promote a culture of accountability, AI strategies should include a clear timeline, with well-defined deadlines and set periods for evaluating progress. Having a clear timeline also encourages governments to factor feasibility into their plans, and to think about what they can realistically accomplish in a set period of time. For instance, the Colombian AI strategy very clearly proposes 7 stages of implementation, with a report produced at every stage of the process reflecting on progress.

3. Measurable goals

By the same token, national AI strategies should always be based upon a series of measurable outputs by which governments can monitor progress and hold themselves accountable to change. This might include anything from tracking the amount of students trained on AI educational programmes, to monitoring savings generated by streamlining government services using AI. 

Governments might look to Uruguay as a good example of this; alongside their AI strategy, the Uruguay Digital website allows visitors to track the progress of initiatives in real time.

Source: https://uruguaydigital.gub.uy

4. Funding

Digital transformation requires significant financial investment. If governments are to benefit from the advantages AI can bring, they should be sure to dedicate a substantial amount of funding into implementing a comprehensive AI strategy. As it stands, a number of governments have created ambitious AI strategies, positioning themselves as digitally innovative administrations, but without offering the funding to back up their plans.

The Chinese government offers an example of how significant funding can lead to important gains in AI. Exact figures for Chinese R&D expenditure are hard to come by, but a conservative estimate would suggest about $138 billion in non-defence funding for AI. This surge in government funding since 2016 seems to have supported an explosion in AI startups in the country.  

Realistically, however, the amount of money that different countries can dedicate to an AI strategy will vary radically. Therefore, international bodies, such as the World Bank or the European Union, also play an important role in ensuring that governments are able to invest in a national AI strategy. Governments and external investors alike should bear in mind that the initial funding needed to design and implement a national AI strategy will pay dividends in the long term.

5. Strong coordination 

A successful national AI strategy requires strong coordination between all of the entities involved in its implementation. The benefits of AI are spread across several sectors, meaning that when governments are designing a broader national strategy, they need to promote a transversal approach which involves deep collaboration between all their ministries, departments, or partner institutions. This might involve the creation of a central unit, like the United Kingdom’s Office for AI, which can advocate for implementation of an AI strategy across government departments.

Similarly, AI strategies need to be responsive to the interests of national stakeholders across civil society and the private sector. In order to ensure that these voices are heard, governments should explore the possibility of carrying out a consultation process whilst designing and implementing an AI strategy, which allows onlookers to provide their feedback.

Over the next few weeks, we’ll be putting these ideas to the test by taking a closer look at a number of national AI strategies, and discussing their strengths and weaknesses. If you have any thoughts about what makes a good strategy, or suggestions for strategies we should consider, don’t hesitate to get in touch – email info@oxfordinsights.com or find us on Twitter, @oxfordinsights