Executive Briefing – AI Opportunities Action Plan Forum 01
Date: 28 January 2025
Location: Phoenix Court, 2 Brill Place, London
Hosted By: Evident, Next Economy Trust, Phoenix Court Works, and partners
The inaugural AI Forum convened a diverse group of stakeholders to support, challenge and track progress of the Government’s AI Opportunities Action Plan. The discussions were broken into six key workstreams reflecting the Plan: infrastructure, data, talent and skills, “go-to-market” (procurement & adoption), and “AI makers, not takers” (building national champions).
This is an executive briefing detailing the key insights, challenges, and proposed solutions across each of these workstreams.
I. Infrastructure
Key Insights:
Infrastructure development must benefit the entire UK, not just the Southeast. There's a need to create infrastructure that connects to areas of need across the country. This requires partnerships with local authorities, universities, and innovation clusters across the UK.
Startups require support to access compute resources, not just academic researchers and support.
There needs to be a long-term plan for compute infrastructure that isn't easily disrupted by changes in government.
There is an opportunity to build data centers on old industrial land, potentially on the East Coast of the UK, to take advantage of wind power.
Challenges Identified:
There's a lack of clarity on how small companies can access resources. This is partly due to research councils not allowing companies access to resources.
The government has only identified one of five AI growth areas and may not have a clear understanding of the key factors needed to make them successful.
The government needs to move away from reliance on external providers like AWS.
The UK is not competitive in the AI space and may be losing out to other countries.
Proposed Solutions:
Rapid evidence reviews to guide the design of growth zones. This review should look at what works and where.
Provide subsidised or free GPU credits to startups. This could be done by having the government underwrite compute resources, allowing infrastructure players to raise capital.
The government should move away from AWS and other similar providers and create its own sovereign compute capacity.
II. Data
Key Insights:
The focus should be on identifying high-value datasets in the public sector and making them accessible. This means treating data as an asset and making it easier to share.
It's crucial to incentivise data sharing across sectors.
Combining datasets can create more value than the data in isolation.
The government should not try to solve the problem, but rather build out enablers that let others solve problems using data.
The value of data is created by solutions developed from that data, not from the data itself.
Challenges Identified:
There's a lack of clarity on how to prioritise datasets. The term "high value" is not well-defined.
Public distrust of the government's use of data.
The government struggles to create multidisciplinary teams with the necessary skills to manage data effectively.
There is friction in government for enabling data sharing, even between departments like the NHS and DHSC.
Proposed Solutions:
Appoint a Head of the National Data Library who can assemble a multidisciplinary team.
Focus on datasets that enable value for different solutions.
Create a marketplace for data licensing to help entities clear rights and license their data for profit. This will help to unlock data from private archives.
Start with small projects that show value, then expand.
Prioritise data sharing for small, impactful projects and demonstrate their value.
Government needs to make data audits public and ask for input from researchers and businesses.
III. Regulation & Governance
Key Insights:
The UK has the opportunity to lead on AI governance.
There needs to be a balance between innovation and safety in AI.
The UK is well-placed to influence larger players like the EU and US.
The UK should not try to be the most innovative, but instead recognise that it is small and focus on collaboration.
Challenges Identified:
Over-regulation could stifle innovation in AI.
There is public distrust of AI.
Firms are struggling with the risks of AI and how to implement it.
Proposed Solutions:
Focus on transparency and collaboration in the AI industry.
Encourage AI startups with a focus on cybersecurity.
Prioritise consumer protection in AI.
Continue to support the AI Safety Institute.
The UK needs to be more thoughtful in how AI is regulated and not just consider it as a single entity.
Provide sandboxes to support AI companies in navigating regulation.
IV. Talent & Skills
Key Insights:
The UK has a relatively good foundation of AI skills.
Need whole-life learning, from primary education to reskilling.
AI skills not just about computer science, but about broader literacy.
AI could be used to enhance education, with AI tutors for teachers and students.
It is important to engage parents and help children to engage with coding and AI.
Challenges Identified:
The specific goals of the skills section in the Plan are unclear.
The current approach is too academic and does not include "learning by doing".
There is a gap between ambition and the current funding.
Current initiatives are piecemeal and not holistic.
Many teachers are not aware of how quickly AI is advancing.
Students who use AI tools may be seen as "cheating," even if they're enhancing learning.
Proposed Solutions:
Create a National AI Skills Academy to ensure whole-life learning.
Focus on making AI ubiquitous and creating a broad base of skills across society.
Use AI tutors to improve learning outcomes.
Partner with other countries to increase the UK's stature in AI.
Use a micro:bit style initiative to get children into coding.
Baseline student performance against current tests and then measure how much the use of AI improves that performance.
Engage with qualification agencies like the GCEQ to make sure there is a computer science component to curricula.
V. Go-to-Market (GTM)
Key Insights:
The focus should be on delivering good outcomes and user experiences.
The government should focus on creating frameworks that enable businesses to identify where the highest value data sets are.
Challenges Identified:
There is a disconnect between the needs of end-users and the way that AI is being developed.
Companies are unsure about the risks associated with AI.
Procurement processes are risk-averse, slow, and not performative.
Startups struggle to navigate the procurement process.
Proposed Solutions:
Use a "test and scale" approach to identify successful use cases.
Develop a framework for thinking about value, with a focus on revenue, efficiency, outcomes, and risk management.
Focus on specific verticals that can leverage LLMs.
Help companies quantify the benefits of AI investments.
Encourage more collaboration before tender documents are created.
Provide assistance to companies to meet security standards.
Government should procure at speed, which means taking more risks.
VI. Makers Not Takers
Key Insights:
Procurement is an often-overlooked lever for creating national champions in AI.
Challenges Identified:
There needs to be a balance between government investment and the freedom of companies to grow.
There is a need to understand the current government landscape (the "as is") before implementing new AI plans.
Government is often not best placed to identify promising technologies.
Proposed Solutions:
Use government funds to drive capital into homegrown venture capital funds — eg via the ECF.
Link government orders with taking a stake in companies.
Streamline procurement processes and give startups more time to respond.
Develop an open-source strategy for industrial AI development.
Consider a "Manhattan Project for AI"
Next steps
The AI Forum will continue to support, challenge and track progress of the Government’s AI Opportunities Action Plan in the months and years ahead. Gatherings will be held quarterly over the next 2 years across the UK, with tentative dates as below:
2025
Wed 30 April
Tues 1 July
Wed 19 November
2026
Tues 20th Jan
Tues 21st April
Tues 7th July
Tues 3rd November