One AI lab. 40 researchers. A wake-up call for every boardroom.

Friends,

Your weekly AI briefing is here - designed to help you respond to AI, not react to the noise. No curveballs. No chaos. Just clarity.

A quick note from Hugo:

I had the chance to spend time with two amazing companies this week, raising our overall AI ambition. Both teams were surprised that the workshop didn't focus on AI. It focused on what problem they should be solving - and how to solve it with superpowers: telepathy, teleportation, time travel. They came away fired up with momentum and a plan.

If you want your team to think bigger on how AI creates impact, get in touch.

Shoutout to my friend John Mulholland for his awesome superpower framing.

📰 This was the week that was...

The Global South stepped up to the microphone - and the world listened.

India hosted the AI Impact Summit in New Delhi - the first global AI summit held in the Global South - bringing together delegates from over 100 countries with a clear message: the future of AI will not be written exclusively in Washington and Beijing. 88 countries signed a declaration committing to inclusive AI development. Meanwhile, Africa staked its claim in the global AI race, with the African Development Bank mapping a pathway to $1 trillion in AI-driven GDP by 2035.

The standout moment? Indian AI lab Sarvam AI launched two foundation models - 30 billion and 105 billion parameters - trained from scratch on Indian infrastructure with just 40 researchers. Built for 22 Indian languages. Sovereign by design. Frugal by philosophy. If you still think AI transformation requires Silicon Valley budgets, this week was your wake-up call.

Let's get into it.

🔥 URGENT PRIORITIES

✅ No fires to fight

✅ Systems stable

✅ Roadmap unchanged

Enjoy the clarity. It's a gift.

🎯 STRATEGIC INSIGHT

The economy is shifting from doing to being - and the Global South may be better prepared than you think

Tension: Western businesses assume AI transformation is a technology race won by whoever spends the most. But the biggest AI stories this week came from places with a fraction of the budget - and arguably more motivation. India and Africa face superhuman-scale problems every day: healthcare for 1.4 billion people, multilingual populations, stretched public services. When you face problems that large, with tools that are suddenly accessible, motivation compounds faster than compute.

Optimistic insight: AI democratises. Someone with a big problem in front of them and a tool to solve it has motivation in a way that someone who is mildly irritated simply cannot match. Sarvam AI proved this week that 40 researchers can rival what thousands achieved elsewhere - building sovereign, multilingual models on subsidised government GPUs. The name "Sarvam" means "for all." That's the design principle, not the marketing.

What's really shifting: The frugal AI movement is gathering serious momentum. A compelling article from Cambridge Judge Business School's CTO argues that most enterprise AI failures are caused by building more than your data, workflows, and governance can actually support. Frugal AI asks: what is the lightest, safest, most maintainable system that can reliably improve this outcome within our real constraints? That question applies whether you're in New Delhi or Northampton.

Why this matters now: As the economy moves from doing to being, populations with a culture of developing "being skills" - meditation, mindset, community over self - may find an unexpected advantage. India's AI summit was structured around "Sutras" (guiding principles) and "Chakras" (areas of action). That language isn't decoration. It reflects a fundamentally different orientation to technology: purpose-first, people-centred, planet-aware. UK leaders should take note.

Actionable takeaway: Apply the frugal AI test to your next AI initiative. Before choosing a model or vendor, ask: what is the minimum data, the simplest architecture, and the lightest governance that can solve this problem? Start there. Scale only when the previous stage has paid for itself.

🤓 GEEK OUT

1) Sarvam AI - India's sovereign AI lab did a DeepSeek on a shoestring

Sarvam AI launched two large language models (30B and 105B parameters) at the India AI Impact Summit, trained from scratch for 22 Indian languages using a team of just 40 researchers and government-subsidised GPUs. The models use mixture-of-experts architecture to keep inference costs low. The company also launched speech-to-text, text-to-speech, and vision models alongside smart glasses called Kaze.

👉 Read more

Why it matters: This is the frugal AI story of the year so far. It proves you don't need billions to build meaningful AI. The strategic lesson for UK leaders: sovereign, purpose-built models designed for specific populations can outperform general-purpose giants on real-world tasks. Think about what a purpose-built model could do for your sector.

2) Frontline Labs - one person building population-level NHS technology

Frontline Labs is a UK-based project developed by Chris Barry with an extraordinary mission: build the technology to run the NHS so it can be given away for free. One person, no external investment, delivering mission-critical digital infrastructure - rapid stabilisation, data platforms, and AI systems - for NHS trusts and public services. They report a 72-hour average time to first deploy and 99.9% platform uptime.

👉 See the project

Why it matters: This is what happens when AI tools are genuinely democratised. One confident practitioner with the right tools can build at population scale. If you think you need a massive team and budget to transform your operations, this is the counter-example. The barrier to entry for building serious technology has collapsed.

3) AWS tackles COBOL modernisation with AI - and shares what actually works

AWS published a detailed blog post this week on real-world learnings from COBOL modernisation. Their key finding: AI is a genuine accelerator, but only when the reverse-engineering phase is done properly first. BMW Group cut testing time by 75%. Fiserv completed a 29-month project in 17 months. Itau slashed discovery and testing time by over 90%.

👉 Read the AWS blog

Why it matters: COBOL still processes 95% of US ATM transactions and runs 43% of global banking systems. The average COBOL developer is 55, and 10% retire every year. If your organisation still runs mainframe systems, AI-powered modernisation just became viable at scale. The clock is ticking on legacy infrastructure - and the people who understand it are leaving.

🎨 WEEKEND PLAYGROUND

🎨 Turn your words into diagrams with Napkin

Head to Napkin.ai and paste in any text - a strategy doc, a meeting note, a half-formed idea. Napkin will turn it into a clean, shareable visual diagram in seconds. They've recently added much better controls over the look and feel of the output.

Perfect for that Monday morning when you need to explain something complicated to your board and haven't got time to open PowerPoint. Paste, tweak, share. Done.

📢 SHARE THE OPTIMISM

If The AI Optimist helps you think more clearly, forward it to someone else navigating the shift.

If it's not quite landing, hit reply and let me know - I read every message.

Stay strategic, stay generous.

Hugo & Ben