Putting AI to Work: Solving the Productivity Challenge

Contents

Executive Summary

This panel discussion frames AI adoption—not frontier AI development—as the critical policy challenge for unlocking productivity gains globally. Drawing from historical analysis of general-purpose technologies (steam, electricity, computing), speakers argue that countries succeeding in diffusing AI across their entire economies will capture the greatest benefits, and that governments have a distinct role in accelerating this adoption through infrastructure, skills, and policy environments that don't inhibit innovation.

Key Takeaways

Adoption ≠ Innovation: The policy focus should shift from frontier AI R&D to economy-wide diffusion. Nations win by scaling adoption across all sectors and firm sizes, not by having the best researchers.
Government Builds Infrastructure, Not Solutions: Effective policy provides skills programs, compute access, cloud platforms, data commons, and clear regulatory guardrails—but doesn't pick winners or mandate specific AI use cases.
Culture and Comfort are Scalable Levers: Individual and organizational adoption accelerates when people have: (a) exposure to successful use cases from sector peers, (b) safe spaces to experiment and fail, (c) business-focused training (not just technical), and (d) leadership championing adoption.
Global South Has an Asymmetric Advantage: Younger, digitally-native populations and less institutional legacy can leapfrog heavy infrastructure by building efficient, localized AI solutions. This represents a genuine competitive opportunity if policy supports rather than restricts.
The Next 5–10 Years Are Critical: The pace of AI diffusion is faster than prior GPTs. Governments that act now on skills, infrastructure, and policy environment will capture disproportionate gains; those that delay or over-regulate risk being left behind.

Key Topics Covered

AI as a General-Purpose Technology (GPT): Historical precedent from industrial revolutions and the centrality of diffusion, not invention, to national success
The Adoption Imperative: Why widespread, economy-wide AI adoption—not cutting-edge R&D—should be the strategic policy focus
Skills & Education Infrastructure: Role of government-backed training programs, early exposure in schools, and upskilling workers across sectors
Public Infrastructure for AI: Compute capacity, cloud services, datasets, and technical standards as government-built "roads and bridges"
Geographic Distribution & Regional Equity: How to spread AI benefits beyond tech hubs (e.g., UK's "AI growth zones" outside London)
Digital Sovereignty vs. Innovation Tension: Balancing data security/control with open borders and free flow of information
Organizational & Cultural Change: Barriers to adoption within companies; need for comfort, experimentation, and iterative adoption
Small Language Models & Localization: Opportunity for Global South to leapfrog using efficient, locally-relevant AI rather than only scaling large models
Private Sector Leadership: Role of companies (Amazon, LocoBuzz) as primary adopters and innovators; government as enabler, not blocker
AI for Development: Emerging focus on using AI to serve developing nations and global majority contexts

Key Points & Insights

Historical Pattern: Diffusion Wins Wars: Past industrial revolutions were won not by the nation that invented the technology, but the nation that diffused it fastest and broadest across its economy. Electrification took 30–40 years to transform the US economy because organizational adaptation was required across many companies and sectors, not just technical deployment.
Speed of Diffusion is Different: Unlike prior GPTs (which took decades to spread), AI is diffusing much faster, creating both urgency and opportunity. However, this doesn't mean adoption happens automatically—intentional policy and private sector work are required.
Government's Role is Infrastructure, Not Direction: Effective government action focuses on: skills training, compute infrastructure, cloud access for SMEs, common datasets, technical standards, and fostering a pro-innovation regulatory environment. Government should "build the roads and bridges," not mandate how they're used.
Regulation as a Bottleneck: Over-hasty or overly restrictive regulation (cited in parts of the US and Europe) is inhibiting adoption. Some jurisdictions are experiencing "buyer's remorse" and reconsidering strict early policies. Risk-based, proportionate regulation is more effective than precautionary approaches.
Adoption Requires Organizational Adaptation, Not Just Technical Capability: Companies adopting AI need to: (a) identify use cases that align with existing tech stacks and workflows, not force-fit; (b) translate AI metrics (accuracy %) into business metrics (ROI); (c) move incrementally from "AI-assisted" to "AI-powered" rather than disruptive overhauls; (d) foster a culture of experimentation and tolerance for mistakes.
Skills Training is Multi-Level and Evolving: Critical needs span schools (teaching logic, information literacy, critical thinking), early-career workers (fluency with AI tools), and mid-career professionals (becoming "managers of AI agents"). The user interface shift to natural language means non-CS experts can adopt tools—but they still need judgment and verification skills.
Small & Medium Enterprises (SMEs) Face Information & Access Barriers: Examples cited (Singapore's "fractional CTO for hire," India's digital public infrastructure) show government can reduce barriers by connecting SMEs with technical advisory, cloud access, and use-case registers. This is especially critical in Global South contexts.
Digital Sovereignty ≠ Isolation: Framing sovereignty correctly—as customer data control, encryption, and transparency—rather than nationalist technology isolation, allows countries to attract investment and innovation while protecting legitimate security interests. Amazon's approach exemplifies this: customers control data location, access, encryption keys.
Geographic Distribution is Feasible: Unlike prior software/SaaS waves concentrated in urban tech hubs, AI infrastructure (compute, energy, data centers) can be distributed. UK's investment in South Wales, North Wales, Scotland, and the Northeast shows how large capital infrastructure can drive regional ecosystems if paired with local skill-building and startup support.
Global South Opportunity: Leapfrogging via Efficiency & Localization: Developing nations can skip building heavy infrastructure for large language models and instead focus on open-source, locally-adapted small language models and sector-specific solutions. This is faster, cheaper, and better aligned with existing workflows than importing large-model infrastructure designed for wealthy markets.

Notable Quotes or Statements

"Success was not about which country produced the latest and greatest new to the world innovations. Instead, success was about which country could win the diffusion marathon." — Jeff Ding (George Washington University, Technology and the Rise of Great Powers)

"The first thing governments can do is not stop [private innovation]. Don't do harm." — David Sapolski (Amazon Chief Global Affairs Officer)

"It's not just a blackbox data center in your community. Part of that compute capacity will be reserved for students in those schools to be able to do experiments that ultimately change their sense of what is possible in their future." — Kisha Nyon (UK AI Minister, on regional AI growth zones)

"Adoption will not happen very drastically. It is a small transformation where it starts from AI-assisted and actually then becomes AI-powered. So it's a journey." — Shuby Agarwal (LocoBuzz co-founder)

"One of the things that generative AI is doing, it is giving us computers that people can talk to in their natural language… It's a very interesting user interface change in the way that people interact with technology systems." — Kevin Allison (Manurva Technology Futures), citing CEO of Cohere

"If you believe that the widespread adoption of economically useful AI applications is important for your country, and if that is a strategic goal, then what is the role for public policy?" — Kevin Allison (framing the core policy question)

Speakers & Organizations Mentioned

Role	Name	Organization
Moderator	Kevin Allison	Manurva Technology Futures (Washington DC, geopolitical/AI policy)
Video Speaker	Jeff Ding	George Washington University (Political Science, author Technology and the Rise of Great Powers)
Panel: Corporate	David Sapolski	Amazon (Chief Global Affairs Officer & Legal Officer)
Panel: Startup	Shuby Agarwal	LocoBuzz (co-founder & COO; Mumbai-based AI customer experience platform)
Panel: Government UK	Kisha Nyon	UK Government, Minister for AI
Panel: Government India	(Referenced but not named in transcript)	Indian Government (Ministry of Commerce/METI representative)

Other Organizations/Initiatives Referenced:

Singapore (digital public infrastructure, "fractional CTO" programs)
IRCTC (Indian Railways; use case for AI social listening)
Cohere (Canadian AI startup)
Sage (UK financial services firm)
UK AI Security Institute
UK AI Incubator
India's Future Skills Initiative
Digital Public Infrastructure (India's cloud services)

Technical Concepts & Resources

Concept	Definition / Reference
General-Purpose Technology (GPT)	Technologies (steam, electricity, computing, now AI) with broad applicability across sectors; require diffusion infrastructure and skills to unlock productivity gains
Large Language Models (LLMs)	Frontier AI models (e.g., ChatGPT, Claude); referenced as globally scalable but computationally expensive
Small Language Models (SLMs)	Efficient, locally-adapted AI models; identified as opportunity for Global South adoption without heavy infrastructure investment
Natural Language Interface	User interface shift enabling non-specialists to interact with AI via conversational language rather than code
Digital Sovereignty	Customer control over data location, access, encryption, and key management; framed as distinct from isolationist technology nationalism
Cloud Infrastructure	AWS, digital public infrastructure (India); cited as critical government-provided infrastructure for SME access to AI tools
Chatbot Window / Chat Interface	Early consumer AI interaction model (ChatGPT, Claude); audience survey shows shift toward code-based tools (Claude Code, Amazon Bedrock) for professional use
Technical Standards & Interoperability	Government role in creating standards allowing data/models from different companies/vendors to work together
Compute Capacity Distribution	UK investing in regional data centers (South Wales, North Wales, Scotland, Northeast) to decentralize AI infrastructure benefits
Open Source Infrastructure	UK positioning as "home of open-source talent"; government support for shareable, non-proprietary AI tools
AI Growth Zones	UK policy initiative creating regional compute clusters paired with local skill-building and startup ecosystems

Additional Context

Event Setting: AI Summit in India (2024/2025); part of broader "AI Adoption Initiative" with workshops in 10 cities across 5 continents
Chatham House Rule: Applicable; information is attributable to discussions but not to specific speakers
Audience: Mix of AI professionals, policymakers, corporate leaders, startup founders; temperature check showed majority using AI daily, but still in early experimentation phase relative to future potential
Follow-up Convening: Extended discussion planned at Meridian Hotel (informal setting) for continued expert dialogue on summit themes