The Raven's Blog

Artificial intelligence: Is it really worth it?

Hugo D’Ulisse, technical director, public sector at SAS UK & Ireland, outlines the top issues you’ll need to consider for making artificial intelligence (AI) a success.

When it comes to AI, it feels like the UK has entered into a sprint to the top. In the adjacent lanes are China and the US with all their spending fire power, followed by Canada, France and Germany. And you can be fairly certain that all other governments in the G8 and Europe have significant ambitions to enter the fray. Is there room? Yes, but each may need to carve out some form of world domination – or at least niche competitive advantage.      

Ambition is a wonderful thing, but only when it’s targeted at the right objectives. Which begs this question, is AI really right and necessary for UK Government organisations – whether central or local? In short, is AI really worth it? By that we mean, will it really have a positive effect? To help answer that question, MIT-Sloane produced a great piece of research that looks at the impact of AI on organisations’ offerings and processes. The results are displayed in graphical form below.

Unsurprisingly, industries such as investment banking, consumer services, insurance and logistics are ahead of the game today, and there’s no real movement in that ranking within the next five years. As you might expect, perhaps due to time, budget and skills limitations, public sector organisations sit at the back of the queue when it comes to the AI effect – both now and tomorrow – but that does not mean that the use cases are not profoundly powerful. They certainly are.

Spanning everything from supporting early cancer diagnosis to online chat bots, identifying real and emerging terrorist threats and helping to reapportion financial resources to more citizens through the prevention of all manner of fraud, AI has the potential to utterly transform outcomes for citizens. I would argue that the use cases of AI for government are at least as powerful as they are in the commercial sector.

The challenge for public sector organisations that are (understandably) playing catch up is, how can they ‘get AI right?’ How can they avoid the pitfalls and accelerate adoption in order that citizens can rapidly experience the value and benefit from substantially improved outcomes? Quickly. 

The path to success: a five-point plan for government

1. Give your objectives a 360-degree interrogation
   Begin with an utterly ruthless, objective mindset. The prospect of being first to ‘market’ with a radical new use case is incredibly seductive, especially in a climate where every mention of AI innovation seems to garner heavy PR attention. However, it’s vital that you thoroughly review all use cases, by involving key stakeholders (without getting bogged down in a protracted process that is unnecessary for reasons we shall discuss in a moment). Having a sketchy understanding of the ultimate outcome(s) you want to achieve will often result in expensive mistakes. Aligning your use cases to your departmental plan sounds like an obvious thing to say, but it’s worth taking your time at this stage. Consider where AI will have most effect. Perhaps, like many commercial organisations before you, it would be most valuable to deploy it to improve service delivery, changing existing internal processes can often be more complex than designing new offerings driven by AI. Perhaps, changing the way you conduct internal processes will help you cut costs in order to prioritise budget Whatever your choice, test, test and test against your strategy before implementation.      
2. Make the most of your data
   This step is very simple. Know your data inside and out. This is because AI was made to ‘learn’ from big data. Factor in citizen sensitivities around the use of their data and how trusting they are with government organisations. Unless you have big enough data sets to work with and the security in place to allow access to the right people, at the right time, you might want to hang fire.    
3. Develop an ethical use strategy
   This is an extraordinarily hot topic and, in my opinion, one that is likely to intensify. In particular, the issue of bias is of great concern to industry experts, the press and, of course citizens, who are concerned that decisions about them will be made by machines built with the implicit bias of the developers. And as algorithms increase in sophistication and ‘learn’ how will you continue to prove they are using data ethically and working without bias. You may be required to prove this at any time! The Future of Life Institute has some very useful food for thought on this matter. From an analytics perspective, it is important that you understand how ‘black box’ solutions prevent you from having full transparency of the algorithms. Open source makes auditing and standardising of development difficult with little auditability built in.
4. Take a fail fast, win fast development approach
   Given what I discussed in the first point, you might think this is a strange thing to say. However, once you understand the strategic value of the use case, development should take an agile, experimental approach, where the learnings from failures are incredibly valuable and should be quickly integrated into positive moves forward. This is where ensuring you have the right skills in place is absolutely critical – skills that range from experienced data scientists to business analysts who can translate the needs of your organisation in analytical requirements and back again.
5. How can you get started quickly with minimal risk?
   AI presents such an incredibly exciting opportunity. The trick is to translate the possibility into something meaningful for the organisations and citizens you serve. And if, as the saying goes, “it takes a village to raise a child,” why wouldn’t you invite lots of skills and experience to help bring complex AI implementations to fruition. You would benefit from significant experience in a much more cost-effective, focused way. Partnership is key because you will need to design and build a cohesive AI development platform and process that can be complicated to plan and operate without proven processes, technologies and experience.

At SAS we are helping government departments, not just in the UK, but in those countries that have also staked a claim to being leaders in the field of AI realise their ambitions. And to do so very quickly and in meaningful and creative ways. After all, we’ve been working in machine learning for more than 40 years, so it’s far from new territory for us. If you would like to know more about our approach, our partnerships with government or more about how we put the intelligence in AI for government, please review the resources available at the SAS Public Sector site.

For more information

Please click here or reach out to me and my team at [email protected]