Covid-19 required government departments to pull off a rapid shift to digital public service provision at the same time as managing a public health crisis. Many turned to robotic process automation (RPA) to quickly redesign their business process and tackle the administrational backlogs that accumulated as workers were confined to their homes.
Local authorities including Liverpool City Council and London Borough of Redbridge recently shared how RPA allowed them to accelerate admin processes, freeing up staff to focus on more important work.
Having accelerated their adoption of RPA, some government organisations are now looking to the next step: adding AI to their automated processes, also known as intelligent automation (IA). Last year, for example, NHS Shared Business Services opened a £250m framework tender for IA services, by which it means a “combination of some or all of artificial intelligence, robotic process automation (RPA) and machine learning to streamline and improve an organisation’s decision making”.
Some organisations have already had some success injecting AI and machine learning into their process automation initiatives. But a recent report by techUK, the trade body for technology suppliers, warned that a lack of skills and strategy could hinder success.
Benefits of intelligent automation to the public sector
The ‘intelligence’ that IA adds to conventional RPA is in many cases the ability to process unstructured data: scanning images, understanding text, perhaps even processing audio. This makes it well-suited to public sector applications, according to Darrell West, director of governance studies at policy think tank The Brookings Institution. “With a lot of public data being unstructured in nature, IA is well-suited to make sense of text or image information that does not have uniform formatting or comes without much organisation,” West wrote last year.
In techUK’s report, it identified five drivers for intelligent automation in government. The first is the pandemic: “Covid-19 has brought into sharp focus the need for the public sector to be able to operate differently – both for the citizens it serves, and for the people that work within it.”
The second is legacy IT, a much-lamented barrier to digital innovation in the public sector. “IA can help to exploit legacy IT, clearing backlogs and helping the infrastructure adapt to new challenges,” techUK says. “Indeed, if systems will be difficult to transform in the medium-long term, using the wrap-around approach will be a more valuable way forward.”
Other drivers include “freeing the workforce to focus on creative solutions and customer needs”; providing ‘citizen-centric services’ by improving the quality of data and automated decisions; and value for money.
UK public sector bodies that have implemented IA include Scottish healthcare trust NHS Lothian, which has used a combination of AI and RPA to process referrals to its gastroenterology department.
The referral and intelligent triage (RITA) system uses natural language processing (NLP) to extract the meaning from referral letters, having been trained on an archive of 12,000 previous examples. The system was able to automatically triage up to 50% of urgent referrals suspected of being cancer-related and has cut the referral-to-treatment process by up to three days.
Other examples include the National Crime Agency, which implemented IA to process the alerts and requests it receives from Interpol, the international police agency. The NCA receives 360,000 of these a year, according to a case study by IT services provider CGI, which previously were processed manually. An IA system has cut processing time by 79%, saving 28,000 hours of manual processing, CGI says.
Research from the US suggests that IA represents a significant minority of government automation projects. A survey of federal RPA programs in July 2021 found that just under a third (14 out of 44) have incorporated IA elements, including nine that include machine learning, six that include image recognition and five each for chatbots and NLP. The Federal RPA Community of Practice places incorporating IA features at the highest level of its RPA maturity scale.
Barriers to intelligent automation in government
In 2019, the UK government set up an Automation Taskforce to accelerate adoption of RPA and other automation technologies across the government. Despite this and the catalysing impact of the pandemic, techUK stated in its report that “industry consensus is that the initial progress and momentum made by the government in automation has slowed.”
The trade body attributed this to five factors, including a lack of skills and funding, the need for ‘a culture and mindset change’, and what its members see as a “pervasively tactical, rather than strategic, approach to intelligent automation”.
The report also criticised the government’s ‘Do It Yourself’ approach to intelligent automation. Additionally, “techUK members expressed concern around previous failures in adoption due to departments’ desires to build in-house solutions from scratch, which have been compounded by issues such as regular department staff changes and a lack of focus on the ultimate value for the citizen,” the report said (techUK’s members include RPA vendors UiPath and Blue Prism).
Anil Vijayan, partner at IT analyst company Everest Group, says that governments’ hesitation to innovate has slowed IA adoption. “The idea of job losses, whether real or imagined, also prevents an aggressive push towards automation,” he adds.
But IA vendors also need to develop business models that are suited to selling to governments, he says, which means longer sales cycles and an ability to accommodate other nuances.
Furthermore, public bodies’ hesitation to adopt IA may be well-placed, as it presents risks and governance challenges beyond those posed by RPA. In its 2020 report on the State of RPA, the US Federal RPA Community of Practice advised agencies to consider agency ‘security approvals’ and ‘any potential ethical and legal considerations’ before pursuing IA.
Without the urgency of the pandemic, the next phase of public sector process automation may prove to be more gradual than the last.