Salesforce's AI Readiness Index - a useful benchmark for policy formulation to boost AI adoption?

Politicians everywhere are vocal about fostering AI-specific policies to drive growth and demonstrate leadership. But translating these promising visions into practical laws, regulations, and investments is another matter. Salesforce recently commissioned Access Partnership to create an AI-Readiness Index to help jumpstart better conversations between businesses, policymakers, and other stakeholders. It evaluates fifteen metrics for assessing structural factors involved in the broader adoption of AI within government and enterprises.

One essential goal was to tease apart some important factors involved in business-to-business (B2B) aspects, which tend to draw less attention than business-to-consumer (B2B) interactions and associated risks. B2B AI use cases tend to involve more complex use cases and a longer sales cycle; require higher trust, transparency, and accountability; and demand tighter data governance and security measures.

Dr Samantha Torrance, Head of Government Advisory at Access Partnership, which works with governments to design policy and regulation explains the process:

We set out to create an index that basically looks at all of the other sorts of indices and measures that are out there to try and create a composite index that is really sort of the go-to in terms of where the UK is and how well is it doing. So it's a composite indicator made up of or consolidated from other data sources, including those from recognized university think tanks such as Oxford and Tulane University, but also at the UN level, and indicators from WIPO as well. And within the creation of that index, we split it out as a government readiness. So how able is the government to use and create an ecosystem where AI can drive growth and competitiveness and a business readiness? So how able our businesses to drive that AI machine forward and contribute to that growth that that we're looking for?

According to the new metrics, the UK is above average in G7 economies regarding AI adoption. Electronic participation and ICT regulations stand out in government readiness, but the UK needs improvement in fostering emerging technologies and human capital. In business readiness, it excels in adopting emerging technology and technology outputs but could improve in the number of AI startups and venture capital.  Other areas for improvement include supporting small and medium enterprises (SMEs) and scale-ups, fostering partnerships, and upskilling talent.

A big theme is the growing need for education and skills development in the broader context of what AI means in society. Antony Walker, Deputy CEO of techUK, a trade association that fosters collaboration across business, government and stakeholders, notes:

The big bit for us is kind of the up-skilling and re-skilling of the existing workforce. It's an area where businesses, I think, can play a really important role but I think we'd be concerned if government was putting all the emphasis on business there. I think government needs to be thinking about that kind of later life, kind of up-skilling bit.

Metrics

The report observes that no exact measurements of readiness exist specifically for AI. So, the new index relied on the following fifteen proxy indicators for assessing businesses’ ability to drive and sustain the growth of AI and the government’s ability to make AI a key driver of economic growth and competitiveness. Here are the metrics:

Business readiness:

• Companies’ Adoption of Emerging Technologies: Extent companies are adopting AI, robotics, app and web markets, big data analytics, and cloud.
• Business Sophistication: Conduciveness to innovation activity.
• Knowledge and Technology Outputs: Ability to create, impact, and diffuse knowledge.
• Creative Outputs: Ability to create and market innovative physical and digital products.
• Labour-Market Reconfiguration Due to Digital Transformation (“Churn”): Predicted reallocation and displacement of jobs.
• Number of AI Start-Ups: Number of active AI companies.
• Venture Capital Availability & Valuation: Size values and dynamics of VC.

Overall, business readiness is much lower than government readiness for all countries. However, the UK is above average on most and on par for adopting emerging technology and the number of AI startups.

Government readiness:

• Digital Evolution Index: Competition and trust in digital environment.
• Digital Government Score: National digital government readiness and trust.
• E-Participation Index (EPI): Electronic government effectiveness in the delivery of public services.
• Open Government Data Index (OGDI): The availability and usability of government data.
• Human Capital and Research: Government spending and support on skills, training, science, and research.
• H-Index for AI Publications: Citation impact of scientific research.
• ICT Regulation (“Governance” Pillar): Effectiveness of government interventions in promoting participation.
• Government Promotion of Investment in Emerging Technologies: Effectiveness of government policies in promoting investment in AI, robotics, app and web, big data analytics, and cloud.

Areas for improvement

Key recommendations for improvement include:

• Targeted support and incentives for training and development through tax credits and creating playbooks that outline best practices.
• Fostering collaboration between academia, industry, and government through government backed AI-centers.
• Supporting AI courses and curricula across various disciplines.
• Developing adaptive regulatory frameworks for enterprise AI.
• Enhancing open data initiatives through quality standards, education, security measures, linking,
• Enhancing R&D investments to transform lab research into commercial products and tax breaks for research that contribute to innovation goals.

One area that stood out was specific recommendations on adaptive regulatory frameworks. Torrance says the overall objective is to create an overarching framework that individual sector regulators can rely on in determining principles for various industries and sectors. She argues that a principles-based approach can generate better growth opportunities than the very specific rules-based approach adopted by the EU.

Towards this end, the report recommends developing regulations that specifically address the needs and impacts of B2B AI that reflect the lower-risk nature of enterprise AI in contrast with existential risks hyped in the media. This will also require creating a clear definition of actors in the AI supply chain and lifecycle distinct from what’s in place under data privacy laws. These should spell out the roles and responsibilities of AI developers, deployers, and distributors. Other recommendations include defining proportionate and appropriate levels of documentation across the supply chain; requiring humans in the loop for high-risk decisions, and requiring user notification for AI interactions and content.

My take

Developing the first set of metrics to compare AI progress could lead to better conversations between governments, enterprises, academics, and citizens. Thinking about the overall framing of AI progress as part of a system seems much better than throwing billions of dollars at new chip factories or encouraging AI innovators to scrape data at scale.

One thing that stands out is the need to develop a curriculum that makes it easier for everyone to understand how AI affects them personally, along with their families, communities, and jobs. This is not just expecting everyone to master data science and learn Python. It is also about improving literacy to understand how and where it can add value while appreciating how it hallucinates, sows distrust and accentuates inequity.

It also seems important to extend the framing beyond just the AI algorithms and notions like “high-risk.” Over-focusing on the AI algorithms themselves risks glossing over the many ways more automated and faster decision-making processes are already reshaping society for better or worse. And when they do, even low-risk things like hallucinated parking tickets, hate-inducing social media recommendations, or AI content pollution can spiral out of control without easy rectification.