The Week The Giants Agreed to Share

Friends,

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

What happened: The companies racing to dominate AI agreed to share the plumbing - Anthropic, OpenAI, and Block co-founded a new Agentic AI Foundation, with Google, Microsoft, and AWS as supporters. At the same time, Disney and OpenAI made a deal to bring Disney to ChatGPT and Google bought Gemini to iPhones and iPads. And to top it all off AI shopping went mainstream just in time for Christmas. And McDonald's pulled an AI-generated Christmas ad after online backlash.

Thirteen sleeps until Christmas. Parties stacking up. Here's what matters.

Let's get into it.

🔥 URGENT PRIORITIES

✅ No immediate fires to fight
✅ MCP standardisation is good news - reduces lock-in risk
✅ AI visibility in retail needs attention before next Christmas

Use the festive calm for strategic conversations about 2026.

🎯 STRATEGIC CONSIDERATIONS

The Plumbing Just Became a Commons

Anthropic announced this week it's donating the Model Context Protocol (MCP) to a new Agentic AI Foundation under the Linux Foundation. Co-founders: Anthropic, Block, and OpenAI. Supporters: Google, Microsoft, AWS, Cloudflare, Bloomberg.

MCP lets AI agents connect to external tools and data - think USB for AI. In one year: 10,000+ public servers, adopted by ChatGPT, Gemini, Copilot, and Cursor.

When infrastructure becomes shared and standardised, the cost of connecting systems drops toward zero. Competition shifts to what you actually do with those connections. This is the pattern we saw with HTTP, containerisation, and cloud APIs. The plumbing becomes a commons. The value moves elsewhere.

Why this matters: The old question was "which AI ecosystem do we commit to?" The emerging question is "what capabilities do we need, knowing agents will increasingly talk to each other?"

• Interoperability is becoming assumed. Building proprietary integrations may be wasted effort if the standard wins.

• Differentiation shifts. If everyone can connect to everything, your advantage comes from what you connect to and what you do with it.

• Watch the governance. Open standards under the Linux Foundation tend to stick.

The bottom line: When competitors agree to share infrastructure, the real competition is moving elsewhere. The plumbing is becoming commoditised. What you build with it is not.

AI Retail: Three Stories, One Shift

Three Guardian pieces this week capture the same shift from different angles: brands working to be visible to ChatGPT, a hands-on test of AI gift shopping, and Moonpig reporting half its sales now involve AI features.

The visibility question is real. Traffic from AI-powered chat tools is projected to surge 500% year-on-year this season. "What present should I get Dad?" is being typed into chatbots, not search engines. Visibility in AI assistants isn't bought through ads (yet) - it's earned through structured data, quality reviews, and content AI can parse. Traditional SEO is giving way to AEO - answer engine optimisation.

The consumer experience is mixed. AI shopping works for mainstream gifts but struggles with independent brands and defaults to familiar names. It currently amplifies existing market power.

Moonpig shows "AI done right." Half of card purchases involve AI features. Two million AI-generated stickers monthly. Writing assistance. Your handwriting turned into a font. Sales up. Subscriptions up 36.5% year-on-year. They're not using AI to replace human connection - they're using it to enhance it.

Why this matters: AI drops the cost of discovery, matching, and personalisation toward zero. That releases a constraint that's existed forever: the gap between what customers want and what was economically feasible to deliver.

Your move: Don't ask "how can AI replace what humans do?" Ask "how can AI make the human gesture easier, more personal, more meaningful?" That's where the value lies - not in automation for its own sake, but in removing friction from human connection.

Capability Without Culture is Just Expensive Infrastructure

Everyone's talking about AI capabilities. But capability without adoption is just expensive infrastructure. And adoption without cultural readiness is just theatre.

The institutions are catching up. When the Bank of England Governor calls himself an AI optimist, that's not hype - that's a signal for the rest of us to lean in.

Business leaders need to stop treating AI as a tech project and start treating it as an economic realignment. Teams need space to explore, not just mandates to adopt. At this week's AI Night School Live Learning Lab, we walked through turning ChatGPT chats into automated agents using n8n - and watched people grapple with what it means to become technologists.

Just because the technology exists doesn't mean transformation happens overnight. This is people change, not software deployment.

Why this matters now: The next advantage won't come from having the best AI - it will come from having the most adaptive culture.

Your move: Identify one workflow where AI could create genuine value. Then ask: what would need to change culturally for us to actually adopt it? Start there.

🤓 GEEK OUT

🔧 How a ticketing system is the key to AI working

What happened: A design pattern gaining traction structures agentic work into three phases: AI creates a plan, humans review and approve, then AI executes. Anthropic's Boris Cherny notes it can "2-3x success rates" - though as models improve, the boundary of what needs planning keeps shifting.

👉 Read the article

Why this matters: Designing the work is as important as the AI itself. This pattern is becoming standard practice for reliable agentic workflows. Worth understanding if you're moving beyond simple prompts.

📜 Really Simple Licensing for AI Content

What happened: RSL 1.0 (Really Simple Licensing) launched this week as an official standard, backed by 1,500+ media organisations including The Guardian, Reddit, and the AP. It lets publishers embed machine-readable licensing terms into their websites - specifying payment terms and usage restrictions for AI crawlers.

👉 Read the Verge coverage

Why this matters: The economics of the internet is changing. As content creators figure out how to license their work for AI training, new opportunities are emerging alongside the risks. This spec is worth watching.

🍌 Is TOON Really Going to Replace JSON?

What happened: TOON (Token-Oriented Object Notation) is a data format for sending structured data to LLMs. It strips out JSON's curly braces, quotes, and repeated keys, using indentation and CSV-style layout instead. Benchmarks show 30-60% token savings with slightly improved accuracy.

👉 Read the deep dive

Why this matters: As you try to do more complex tasks with AI, condensing token usage becomes a real consideration. I've used TOON in some of my processes and found it very effective - I created a tool to take a data-driven approach to use case discovery, and the volume of data meant I hit the limit of the context window. TOON fixed that.

🎨 WEEKEND PLAYGROUND

🎄 Go to Moonpig.com and see if you can spot the AI

Look for the writing assistant, the AI stickers, the handwriting feature. While you're there, you can tick off one of your Christmas jobs. Two birds, one stone.

📅 UPCOMING

AI for Changemakers - Free Event

📅 Thursday 18th December | 90 minutes | Online

👉 Register on Eventbrite

An interactive session with my friends at Mayvin - a consultancy specialising in Organisation Development and Leadership - on how teams can start experimenting with AI and build the habits to work alongside it. Free to attend and highly recommended if you’re looking at the people change bit of AI adoption.

📢 Share The Signal

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

Not quite landing? Hit reply. We read every message.

Until next week,

Hugo & Ben