AI for bureaucratic productivity: Measuring the potential of AI to help automate 143 million UK government transactions

TL;DR

The size of this opportunity to improve public service productivity through the automation of complex but repetitive bureaucratic tasks, freeing up the time of skilled staff is explored by mapping out the scale of citizen-facing bureaucratic decision-making procedures within UK central government, and measuring their potential for AI-driven automation.

Abstract

There is currently considerable excitement within government about the potential of artificial intelligence to improve public service productivity through the automation of complex but repetitive bureaucratic tasks, freeing up the time of skilled staff. Here, we explore the size of this opportunity, by mapping out the scale of citizen-facing bureaucratic decision-making procedures within UK central government, and measuring their potential for AI-driven automation. We estimate that UK central government conducts approximately one billion citizen-facing transactions per year in the provision of around 400 services, of which approximately 143 million are complex repetitive transactions. We estimate that 84% of these complex transactions are highly automatable, representing a huge potential opportunity: saving even an average of just one minute per complex transaction would save the equivalent of approximately 1,200 person-years of work every year. We also develop a model to estimate the volume of transactions a government service undertakes, providing a way for government to avoid conducting time consuming transaction volume measurements. Finally, we find that there is high turnover in the types of services government provide, meaning that automation efforts should focus on general procedures rather than services themselves which are likely to evolve over time. Overall, our work presents a novel perspective on the structure and functioning of modern government, and how it might evolve in the age of artificial intelligence.

Figures (4)

• Figure 1: Overview of the data collection process and measurement approach. This figure shows a schematic visualisation of the data collection process, including which data sources were accessed and how, alongside an explanation of the AST index and PGenAI rubric. (A) The data collection workflow including the methods of data collection, preprocessing steps, and the two main study questions. The bottom panel plot shows the change in yearly count of number of government organisations registered on .gov.uk and change in total number of .gov.uk websites. (B) The equation for the AST index used to measure automatability of decision-based services, and an example illustration of the task content for the service 'Get a passport urgently'. (C) Distribution of AST index values for decision-based services ($n=201$) including high-volume services graded using PGenAI rubric ($n=20$), squares represent decision-based services and circles represent a subset of high-volume, low-automatable services, respectively (with axis values shown in orange on the left). AST index values of three UK government services included in our sample are highlighted for illustration, measured as of latest year of recorded service existence (2023). (D) An example of the tools and methods of work description for the high-volume service 'Apply for a basic criminal record check', used along with the service task content in assigning a PGenAI rubric score according to the scoring guide, depicted in schematic form in the bottom of the panel.
• Figure 2: Quantifying the scale of UK government and public service transaction dynamics. (A) Distribution of service counts across all government organisations categorised by service topic (numbers in brackets indicate total service count). (B) Structured network of UK government organisations. Node size indicates the count of services offered by a government organisation, arranged from center to periphery according to count of shared service topics between organisations. (C) Distribution of transaction volumes for government organisations ($n=13$) with 10 or more services; remaining organisations are grouped under label 'Other'. (D) Distribution of transaction volumes across all service topics. (E) Distribution of service transaction volumes ranked in descending order, with selected services annotated. A legend is provided at the top of the figure for all panels.
• Figure 3: Measuring the automatability of public services. Distribution of AST index scores across decision-based services (n = 201) grouped by (A) department and (B) topic. Services with an AST index score $\geq 0.75$ (equivalent to a share of routine tasks $\geq 0.75$ ) are considered automatable, as indicated by the dotted black line. The red lines show the mean AST index score for each organisation and topic, respectively. Selected services are annotated.
• Figure 4: Predictors of automatability and estimates of potential benefit from adopting generative AI. (A) Differences in the AST distribution based on differences in service task counts. Whiskers are 1.5 times the interquartile range. The solid black line inside each box indicates the median and the green triangles indicate the mean. (B) Histogram of AST index scores plotted against transaction volumes shown as a percentage of total volume across only decision-based services ($n=201$) and all services ($n=377$). Numbers above bars show count of services for each category. For illustrative purposes, we only show the range [$0.5,1.0$]. (C) Low-automatable decision-based services prioritised by the UK government ($n=20$), defined as having an AST index score $< 0.75$, grouped by potential for benefiting from the adoption of generative AI (Column 1). Column 2 shows the topics associated with the list of services for each segment. Cells are colour coded according to the number of government organisations involved in the delivery of listed services, as per the legend on the top left of the figure.