Using AI to create a more human DWP
Director, AI and Innovation, Department for Work and Pensions
A few months ago, I became the Director of AI and Innovation, charged with helping the Department for Work and Pensions (DWP) to bring together the huge scale of our data and digital systems and our extraordinary people, serving the public day-in and day-out, to improve our services using artificial intelligence.
If we get this right, AI will allow us to create a more human service that cares better for the vulnerable, provides colleagues with more rewarding jobs and takes the pressure off public finances so money can be spent where it matters most. I do not downplay the challenges in achieving this. Issues around privacy, bias, or the impact on how we work are really important, and we need to design with the whole of the service in mind, baking-in safeguards, transparency and intelligibility.
I want to point to a few things that we’ve learned – although please note that we are still learning and need to continue to collaborate and learn from each other!
- No new problems. Technology opens up opportunities to do more. We need to be aware of this and be mindful of our objectives when solving problems. We must demonstrate how AI can be an answer to problems that our colleagues on the front lines already experience. It is essential in my view that we are able to demonstrate what can be achieved today, and involve broad groups of colleagues from across the organisation when creating solutions.
- Focus on root causes of the biggest problems. The range of potential interventions that AI affords, coupled with its high profile, means that there is a very real danger of expending finite resources on lower impact solutions. Through process or task mining we can get closer to understanding where we can prevent problems rather than dealing with their consequences more efficiently. A great example of this is accuracy. If we can give our colleagues tools that assist in getting to the right answer first time, we can forestall the expensive and time-consuming consequences of errors.
- Sharing and openness. If Government is to accelerate the uptake of AI and realise the value it can bring, we need to learn from each other and be able to build on that work. Working in the open is a central part of our approach, including outside of Government. Transparency is the best solution for addressing legitimate concerns about putting smart technology to work on people’s data. While it is not always easy, we need to build a wider compact across civil society on the safe boundaries for using these advanced tools.