There has been a lot written about how AI will change the world of work. Fewer commentators, however, have set out how it might help governments around the world to achieve more for less. A report by Deloitte on that issue makes some very good points.
The potential savings from even simple AI applications could be huge – Deloitte’s analysis suggests that millions of working hours could be freed up by use of AI technology. The range of possible savings is from 96.7 million to 1.2 billion working hours per year, with financial savings of $3.3 billion up to $41.4 billion. That is a lot of additional services that could be provided or government debt that could be paid off.
Cognitive automation can help to overcome resource constraints – Cognitive automation can perform routine tasks at unprecedented speeds, freeing up workers’ time to concentrate on the more difficult aspects. For example, electronic document discovery systems can track down 95% of documentation for court cases, compared with the 50% that people can manage, and in considerably less time.
AI can also be used to reduce backlogs – Part of the problem with government systems is the sheer number of applications and cases that are received. Many are straightforward, however, and could be processed relatively quickly and easily by AI systems, which could flag the more difficult cases for human caseworkers to examine and process.
Cognitive insights via real-time tracking offer security and public health benefits – AI systems using real-time tracking are being used by the US Centers for Disease Control and Prevention to track polio virus. The AI tool can classify virus type, and separate disease reports into relevant clusters, supporting ongoing disease surveillance, and helping to manage outbreaks. This type of pattern recognition is widely applicable for complex problems.
AI is also being used to improve resource usage in food safety and hygiene inspections – The Southern Nevada Health District has started to use an AI application to analyse Tweets reporting food poisoning. The app uses geotagging and natural language processing to identify reports of food poisoning and any associated restaurants. An experiment in Las Vegas found that food safety inspections targeted using the app were significantly more likely to find problems than randomly targeted inspections.
AI chatbots have been used to improve customer responses via cognitive engagement – The US Army uses an AI chatbot called SGT STAR to interact with visitors to its recruitment website. SGT STAR can answer questions from potential recruits about likely specialisms, benefits and commitment, and direct them to a human recruiter if necessary. This is not a matter of replacing a human with a chatbot, but supplementing the human provision, and creating a partnership.
The biggest improvements come from putting multiple capabilities together – There are individual ‘wins’ by applying AI in several possible areas and ways, but the biggest potential lies in combining capabilities. For example, automation can be used to free up time, and then cognitive insights can be used to identify where best to use that additional resource. The focus needs to be on maximising the flow of tasks.
There are four ways to use AI: to relieve, split up, replace and augment capacity – Relieving uses AI to automate routine work, freeing up worker time to do higher value work. Splitting up involves separating jobs into steps or tasks, and automating as many as possible. Replacing takes whole tasks or jobs and replaces them with a machine. All these are effectively ways to increase efficiency by automation, and most aim to reduce costs.
Augmentation creates genuine human–machine partnerships – Augmentation makes workers more effective, by giving them additional support to complement their skills. This is very much a partnership, which is better than either machine or human alone. It is designed to help to address problems that could not be managed before, such as detecting insider trading on stock exchanges, or improving treatment choices in healthcare.
Potential uses of AI should be carefully assessed before adoption – There are no obvious answers to where and how to adopt AI technology in government. Deloitte’s report suggests that agencies and governments should consider whether each option is viable, valuable and/or vital, and make a decision on that basis. Things can, for example, be viable without being valuable, and those where AI is vital offer the biggest potential benefit. Policy-makers need to use this type of decision-making framework to maximise public value for money.