Infinite Digital Public Assistants
New TBI/Faculty AI report on the role of AI assistants in govt, and how to get there in one parliament
I’m currently in Taiwan on sabbatical, but had to quickly hop on here to let people know about a big project I was working on before I went away. Apols for brevity - on a phone here¡ I’m back in the office in a month and am v keen to touch base with folks once I return. Let me know if you are too¡
The broken foundations of our state mean that nothing seems to work in Britain anymore.
In the same week that the British sewage companies polluting Lake Windermere plan to pay £320m in dividends to shareholders, OpenAI and Google DeepMind launch demos for real time multi-modal machines that you can work with. The dissonance of 21st Century Britain is have new virtual assistants that can book you appointments for anything, apart from your GP because waiting times are too long. It does not have to be this way.
But have no fear. In what must be the Primark-Greggs collab of the tech and policy world, a new report from the Tony Blair Institute and Faculty AI presents a compelling vision for 2030, outlining a strategic plan for integrating AI assistants into public services to mend our fractured state.
The report (link here ) proposes a transformative new operating model for the interaction between citizens, civil servants, and ministers with AI technology. By 2030, we envision:
- **Digital Public Assistants**: Every citizen should have access to a digital assistant for tasks like filling out tax returns.
- **AI assistants in Civil Service Teams**: Civil service teams should integrate AI agents to enhance productivity, such as using AI to red team documents.
- **Digital Policy Twins**: Simulating decisions with digital twins to enable safe and efficient policy experimentation.
This vision isn't just theoretical. The report dives deeply into how these changes can be implemented and addresses critical issues around privacy and autonomy.
Beyond improved service quality, we also have calculated this could save HMT £40bn a year. However, the current infrastructure and talent pool are inadequate for this vision.
To deliver on this vision, the UK government must consolidate fragmented government AI efforts under a single entity—AI Mission Control (AIMC). This body should coordinate departmental strategies, test and scale new infrastructure, and reform civil service learning. Additionally, launching an AI talent surge within government departments and establishing fast-track streams for AI expertise, with salaries set at 75% of market rates, is essential.
Implementing these changes is a monumental task, fraught with unpredictable advancements in AI, significant resource demands, and sociotechnical challenges. However, it represents the most promising path to modernizing government delivery models.
A heartfelt thanks to the brilliant minds I worked with on this project, including Alexander Iosad, David, Ursule Katijole, Rachel Irwin, Sona Hathi-Rehman, and Filip Wolski. Your work is paving the way for a more efficient and effective public service landscape.
Two minor points to flag at the end:
I’m keen to write more on these matters (and other sci-tech pol stuff) over the summer, and am keen to place op-eds if there are interested editors
We also welcome feedback on this work. It’s very very ambitious and we need big thinking and big building to work side by side to actually fix our shattered social contract