How AI and ML can help the public sector push boundaries
The UK government recently launched its first National Artificial Intelligence (AI) Strategy to help ‘strengthen its position as a global science superpower and seize the potential of modern technology to improve people’s lives and solve global challenges such as climate change and public health’. In response, public sector agencies are increasingly exploring how AI and machine learning (ML) can help to modernise processes and solve complex problems.
At Kainos, we’ve worked across multiple public sector AI projects to support this aim and help public bodies leverage data analytics to positively impact the lives of millions of citizens. To find out more about how AI and ML are helping to transform public sector, we spoke to Piers Campbell, our Head of Technology – Data & AI Practice, on how AI and ML are driving innovation in the public sector. Piers has over 15 years of experience delivering innovation to customers across multiple sectors around the globe – including government, energy and telecoms – helping them make better use of data and build their AI capability.
What’s your background in ML and how have you seen the technology evolve?
I began my career in data science and ML about 20 years ago. The field was very academic and research-focused back then. But over the last 10 years, we’ve seen huge change in our industry towards applications and use cases. ML has moved from theory into practice. It’s becoming more mainstream and is now part of almost every software system. That change has really accelerated in the last five years, particularly with major players such as AWS making ML much more accessible.
How do you think ML can be best applied to the public sector and can you talk us through some use cases?
The public sector is really embracing the technology and recognising that ML and AI are a gamechanger—it’s marvellous to witness. With ever-tighter budgets, many departments and agencies are using AI and ML to create cost and efficiency savings – freeing up civil servant time and better serving citizens. Any data heavy workflow, or one that requires lots of actions from a human, is a good candidate for AI and ML. As are processes that have a lot of repetitive actions where typically a human is charged with spotting anomalies, like checking a regulatory submission.
For example, Kainos worked with DVSA to pioneer a ML tool that enables DVSA to identify abnormal MOT testing behavioural patterns, ultimately improving Britain’s road safety. The Land Registry (HMLR) is using AI to automate deed comparison when processing planning applications, releasing caseworkers to focus on more valuable work and speeding up the process for applicants. In both cases, we worked with Amazon Web Services (AWS) and leveraged Amazon SageMaker, a flexible cloud Machine Learning platform, to develop and deploy elements of the solution.
What challenges does the public sector face when it comes to applying ML?
Its biggest challenge is attracting and retaining talent. The public sector is recruiting in a seriously limited market. There is a skills shortage - data scientists and AI professionals are in high demand. And that leads to the next challenge, which is that the public sector is competing with private enterprises with far deeper pockets. The public sector often trains ML professionals who then leave for the private sector to make more money. That’s why Kainos works hard to help customers grow skills and knowledge in-house. We don’t just deliver a service; we enable our customers to continue to manage their projects independently, to contribute to a rewarding work environment.
What recommendations would you have for overcoming these challenges?
Several government departments are already strengthening their data and digital functions by offering more competitive salaries and removing some barriers to employment. Many AI professionals want to use their skills altruistically to deliver tangible benefits and make the world a better place. If the public sector positioned AI as a public good, it would likely attract more of the right people.
Widening routes to access jobs in the field and looking at non-traditional entrants is also important – not everyone working in AI needs to have a PhD, or even deep knowledge of how to build an algorithm. AI is becoming more usable in out-of-the-box scenarios, so what we’ll need more of is everyone from philosophers to communicators to more generalised professionals. We also need to look at engaging people in AI and ML far earlier on in school to capture their interest.
Fortunately, we’re now seeing more investment in developing skills and creating a more diverse workforce. For instance the UK government recently announced a £23 million fund to create 2,000 scholarships in AI and data science in England. The money will fund conversion courses to help underrepresented groups get jobs even if they have no previous experience. It’s a great investment in the future of the industry.
What are your thoughts on the UK’s National AI Strategy and what do you think the UK needs to do to make this ambition a reality?
The UK needs to take multiple approaches. Firstly, it needs a mechanism such as a visa system to bring talent in from outside the UK. Ideally, we would build partnerships between our universities and those overseas to make the process smoother. Additionally, as mentioned earlier, we must accelerate our own talent development and create more routes to entry, such as apprenticeships. Here at Kainos this has been a huge driver of our growth and success, and we find a lot of our apprentices stay on for years.
Another vital component is to increase grassroots investment and build IT and ICT into school curricula. That’s why we created the Kainos Academy in 2011. We wanted to tackle the digital skills gap by capturing the imaginations of young people through various events like Kainos Code Camp and BelTech EDU. We also offer an Earn As You Learn programme, where school leavers can get paid, get real-life work experience and study for a degree part-time.
Where do you think the UK sits on the AI ‘superpower’ list and which countries are leading the way?
The UK is probably in the top five or six, but we are behind nations like the US and China. We’re certainly not leading the way. There are little pockets of innovation, but we can do better in those areas we’ve discussed, like skills development and recruitment. Yet it is equally important to have the capability to convert AI research into practice, and I think we still have a gap there too. The UK’s low tolerance for risk means many AI entrepreneurs tend to leave for places like the US, where they can attract higher investment, and where there’s more tolerance for failure. This creates a brain drain.
Our academic record means that we punch way above our weight in AI and ML research, but the focus now needs to be on embedding what we’ve learned into practical applications. Though people can be quick to criticise public-private partnerships, they can be very beneficial. One example that stands out is research institutions, such as the Turing Institute, which have shown that combining public sector budgets with a private sector laser focus can deliver tangible results.
What do you think the role of regulation and government will be in the future to deal with ethical issues and challenges of bias in AI?
Ethics is certainly becoming more of a concern in the field and at Kainos we’re very aware of this. We recently worked with Tortoise and more than 20 leading AI experts to explore the future of trust in AI, and found that we all shared similar concerns – there's a growing demand for professionalisation, standardisation and mechanisms for disclosure within the field.
We believe that all organisations have a responsibility to address these challenges – that's why we produced the report, to advance this conversation around AI ethics. And we’re making sure we’re promoting best practice internally, too - we’ve retained the IEEE as an ethics advisory body, and joined the TechUK Data Analytics & AI Leadership Committee, as well as hiring our own data ethicist. So there is a lot that can be done by individual organisations, outside of formal regulations.
As for formal regulation, I’m not sure that will prove to be the right way. But the industry certainly needs some structure and guidelines, and I’m hoping the Office of AI will provide that. Kainos has been working closely with them so I think this will happen. Our public sector customers want to be proactive and use AI to resolve problems, but there’s also a lot of work to do around educating people on what AI really means.
A question we hear often from customers is how do we overcome natural bias in our data? Datasets often have bias built in and you have to go back quite far to understand what it is and where it came from. Data cleansing is an essential part of any AI or ML project. We work with customers to help them unpick their data and design models that avoid bias.
The UK government has a 10-year plan; where do you see AI and ML in the public sector going over the next decade and is 10 years an achievable timeframe?
A lot can happen in 10 years and the rate of change is accelerating. If we don’t make significant strides in the next decade, the UK will be left behind. For me, it’s about trying to tie all the pieces together. Where can we open opportunities and break down barriers to apply AI? Essentially, AI and ML are about solving problems and wanting to help, and it’s important we maintain that focus.
Piers Campbell, Head of Technology, Kainos