Editor’s note (15 October 2026): Editor’s note, October 2026: this article was written in November 2023, when generative AI in government was at an early stage. Adoption, regulation and Scottish programmes have moved on considerably, so please read it as a historical view.
In short: This 2023 article argues generative AI could reshape government technology, from infrastructure inspection to benefits advice and policy simulation, though early public sector uses mostly refine existing practice. It stresses that bias, misinformation and accountability risks mean ethical, inclusive design must lead, an argument relevant to Scotland’s public sector.
Key takeaways
- Generative AI refers to machine learning models that create realistic synthetic media, content and code from text prompts, including images, poems, music, dialogue and software scripts.
- The post imagines field inspectors using prompts to see visualisations of structural defects, and older constituents asking a voice bot plain questions about policy paperwork.
- The post says global examples so far mostly refine existing practice rather than wholly disrupting governance, though it expects more foundational shifts as models grow and apps simplify delivery.
- Challenges include bias in training data, misinformation spread at scale, and accountability given the opacity of large models, which makes policymakers cautious about a dual-use technology.
- The post concludes that principles of ethical and inclusive development should lead the way, so calls for caution do not prevent seizing opportunities for public good.
An earth-shaking revolution in artificial intelligence is afoot that promises to reshape government technology as profoundly as the web browser or iPhone did decades ago. 2023 appears poised as a breakout year for integrations of generative AI across the public sector tech stack and digital government services in ways that could redefine civic infrastructure everywhere.
What is generative AI?
For the uninitiated, generative AI refers to cutting-edge machine learning models like OpenAI’s DALL-E, Google’s Imagen, and Stability AI’s Stable Diffusion that can create strikingly realistic synthetic media, content and code simply from text prompts. Whether photorealistic images, lyrical poems, musical compositions, natural-sounding dialogues, or software scripts, these algorithms portend automation of creative tasks previously thought the sole domain of humans.
And while debates rightly rage about ethical application and potential misuse, benign deployments tailored towards enhancing public sector missions could reap tremendous upside. As just one example, imagine field inspectors assessing a dodgy railway bridge or corroded gas main feeding prompts like “generate 3D model of structural defects” to instantaneously see detailed visualisations only before accessible via costly engineering analysis or dangerous manual instrumentation.
How could generative AI improve public services?
For serving citizens more intelligently, imagine elderly constituents struggling with convoluted policy paperwork instead asking plainspoken questions to a voice bot that responds conversationally while pulling generated summaries perfectly tailored to circumstance. Or otherwise, siloed agencies combining data assets to power simulated policy scenarios mixing open debate transcripts, academic discourse and historical subdomain reports to better understand impacts of potential legislation before any spending occurs.
Where is generative AI already being used in government?
These may sound like science fiction scenarios with little bearing on practical government operations. But global examples demonstrating in-production value are already emerging:
- Singapore’s government IT agency has developed Project VAI, a proof-of-value prototype using generative models for automated visual inspection analysis of infrastructure like buildings and telecom towers. Rather than relying purely on manual monitoring, prompt-based AI could automatically detect cracks, corrosion or other issues.
- The UK’s Department for Work and Pensions trialled a prototype named Avatar-Assist which provides bespoke explanations to benefits applicants by generating custom responses to posed questions based on analysis of over 24,000 pages of policy guidelines – far more than any human agent could master.
- Delaware’s Office of Management and Budget is actively experimenting with leveraging GPT-3 style models for summarising complex budget allocation decisions and analyses into simplified formats for legislators and public comprehension based on prompts incorporating data-rich funding reports.
These breakthroughs mostly refine existing practices rather than wholly disrupting governance thus far. But make no mistake – more foundational shifts are coming. As models grow ever larger and delivery mechanisms become democratised into simple apps, the technology and economics of generative AI will enable new classes of data-intensive, participatory and experimental policymaking simply not viable through legacy methods. 2023 appears a breakout year.
What are the risks of generative AI in the public sector?
Yet challenges abound, including around bias in training data, perpetuating misinformation at scale, and accountability given black-box opacity of large models. Policymakers are rightly cautious navigating such a dual-use technology. But calls for precaution should not prevent seizing opportunities for public good – provided principles of ethical and inclusive development lead the way.
With responsible design, generative AI could help governments transcend intractable trade-offs between policy sophistication, quality of insight and democratic engagement… heralding a new era where constrained resources never prevent modelling the complex, dynamic issues that define civilisation governance itself – from pandemic preparedness to climate change, systemic inequality and beyond.
2023 stands ripe with potential for vision and leadership to set the trajectory not just for rapid uptake of generative AI in GovTech, but to wholly reimagine what responsible technology augmented governance could resemble across this century and beyond for the benefit of all people. The models may be nascent today but their societal impacts could soon reshape far more than any individual government service alone.
Want to discuss your use of Generative AI in this important sector? Contact us today.
Frequently asked questions
What is generative AI in government?
Generative AI refers to machine learning models that create synthetic media, content and code from text prompts. In government, the post imagines uses such as field inspectors generating visualisations of structural defects, voice bots giving elderly constituents tailored policy summaries, and agencies combining data to simulate policy scenarios before any spending occurs.
What are examples of generative AI in the public sector?
The post cites Singapore’s Project VAI, a proof-of-value prototype for automated visual inspection of infrastructure; the UK Department for Work and Pensions trial of Avatar-Assist, which answers benefit applicants using over 24,000 pages of policy guidance; and Delaware’s Office of Management and Budget, which is experimenting with summarising budget decisions for legislators and the public.
What are the main risks of generative AI for public services?
The post highlights bias in training data, perpetuating misinformation at scale, and accountability given the black-box opacity of large models. Policymakers are described as rightly cautious about such a dual-use technology. The post argues that precaution should not block public benefit, provided ethical and inclusive development principles lead the way.
Will generative AI transform government according to this article?
Not immediately. The post says early breakthroughs mostly refine existing practices rather than wholly disrupting governance. It still expects more foundational shifts as models grow larger and delivery becomes simpler, enabling data-intensive, participatory and experimental policymaking that is not viable with legacy methods, as long as responsible design underpins adoption.