Generative AI: The Savior of Sluggish Government Services or a Pandora's Box of Unintended Consequences?

The Turing Institute's study on the potential of generative AI (GenAI) in government operations provides a compelling analysis of how AI could revolutionize the public sector by automating a vast range of tasks currently performed by civil servants.

The Turing Institute was named after Alan Turing and grew out of the Machine Intelligence Research Unit at the University of Edinburgh. It is now the UK’s national institute for data science and artificial intelligence.

This study underscores the immense potential for efficiency gains, the challenges of implementation, and the broader implications for public service delivery.

In summary, the study identified that:

• The UK government carries out around 1 billion citizen-facing transactions per year across almost 400 services. The study focused on 201 of these services that involve a decision and information exchange, such as registering to vote or applying for a national insurance number.

• These 201 services account for around 143 million complex but repetitive transactions. The study found 84% of these transactions could be easily automated with AI, representing a huge potential opportunity.

• Even if AI could save just 1 minute per transaction, that would be equivalent to hundreds of thousands of hours of labor saved each year according to Jonathan Bright, head of AI for public services at the Turing Institute.

• All 20 services from the DVLA (Driver and Vehicle Licensing Agency) and DVSA (Driver and Vehicle Standards Agency) were found to be easily automated by the study. However, far fewer services could be automated at the Courts or Department for Education.

• The most difficult service to automate was "appeal against a visa or immigration decision" with only 38% of tasks considered routine. Services like this are expected to retain a considerable human component.

• The most easily automated government service topics were areas like "driving and transport" and "training and skills", while the hardest to automate included "benefits", "childcare and parenting", and "national security".

• The study did have limitations, including that it was largely conducted through desk research without collaboration with service providers directly. More granular task-based mapping involving engagement with service delivery professionals should be the next step.

• A separate Turing Institute survey of 938 UK public sector professionals found generative AI use is already widespread - 45% were aware of its use in their area of work and 22% actively use a generative AI system themselves.

Despite the promising potential, the study also highlights several challenges and considerations. The introduction of AI in public services must be approached in a manner that benefits everyone, addressing concerns around job displacement and ensuring that the technology's deployment is responsible and equitable. 

The study's findings may have profound implications for public service delivery and suggests generative AI may have enormous potential to automate a large portion of repetitive government service transactions, but achieving this will require significant work and investment.

Use of the technology is already growing organically in the public sector.

Sources:

[1] https://www.turing.ac.uk/news/publications/response-government-call-evidence-generative-ai

[2] https://aws.amazon.com/institute/demystifying-generative-ai-for-government/

[3] https://www.computerweekly.com/news/366575212/Turing-Institute-Generative-AI-is-going-to-change-the-face-of-government

[4] https://www.turing.ac.uk/news/publications/generative-ai-already-widespread-public-sector

[5] https://www.publictechnology.net/2024/03/25/government-and-politics/more-than-four-in-five-government-transactions-ripe-for-automation-turing-study-finds/

[6] https://www.turing.ac.uk/research/research-programmes/public-policy/public-policy-themes/ai-public-services

[7] https://www.turing.ac.uk/research/research-projects/understanding-use-and-impacts-perceptive-predictive-and-generative-ai

[8] https://www.lexology.com/library/detail.aspx?g=c196cded-b2bc-45e4-a0e9-3e4115744316