The Great AI Expansion

There’s a new kind of momentum building in AI. Not just in model performance, but in how people are thinking, asking, designing. We’re moving beyond the novelty of chatbots and into something more structural, more ambient. AI is no longer something we use. It’s becoming the space we operate in. This shift is subtle but significant. It means the interface is no longer the product. It's the starting point. In this emerging environment, AI isn't just responding; it's connecting, orchestrating, acting.

It’s not just about better models or smarter chat. It’s about integration. Absorption. AI systems are starting to touch everything: inboxes, calendars, meeting notes, file systems, marketing pipelines. And they’re beginning to act on our behalf – not just advise, but plan, decide, do.

A year ago, most of the leaders we were working with were still learning about effective prompting. Now, they’re asking how to build a Chief of Staff. Not a chatbot that mimics one, but something that actually works like one: sees your diary, reads your inbox, understands your role, surfaces the right priorities, and maybe even briefs your team.

Right now, that dream is a little out of reach for most of us. You can cobble together a Custom GPT, sure. But it’s manual. You have to tell it what’s in your calendar. You have to feed it the files. You’re still doing the connecting. It’s not yet acting on its own initiative.

But here’s the thing: you’re close. And getting closer every two months. As Kevin Weil, OpenAI’s Chief Product Officer, put it recently on Lenny’s Podcast:

“Our general mindset is that every two months there’s going to be a better model and it’s going to blow away whatever the current set of limitations are”

Technologies like MCPs (model context protocol) are making it easier for AI to talk to and even use other apps – and it is being put to good use by agentic systems like Manus. They let your AI talk to your calendar, your browser, your CRM – without code or fiddly automation just prompts. Think of it like a Boss AI that tells the other AIs what to do. It can take a vague instruction – “set up a digital twin of our CEO” – and go figure out how to do it. Research the industry, triangulate the role, propose a plan, even build a detailed guidebook.

This is what we mean by “the great AI expansion.” And it’s not just about tools getting smarter. It’s also about us getting smarter about what’s possible. We’re in a state of co-evolution with AI systems. As our literacy improves, we come up with better ideas. As we come up with better ideas, models have to rise to meet them. That demand loop is accelerating the field just as much as model improvements.

This evolution is happening fast. Faster than most teams can keep up with. Because of the two-month innovation cycle. Every 60 days, the landscape shifts. New capabilities appear. Tasks that felt complex last quarter become trivially easy.

That has huge implications for how organisations build.

You have to hold your current ideas and systems lightly. That beautifully crafted prompt flow or custom GPT you just spent three weeks refining? It might be obsolete in a month. And that’s not a failure – it’s a feature of the environment. The work you’re doing now is still valuable. It builds your fluency. But it’s also scaffolding. It might be replaced, but you’ll have the architecture of your thinking to build on.

For example, we’ve been working with an executive team to create realistic digital twins to help them speed up certain tasks. Until recently, this took weeks of effort: prompt engineering, data gathering, persona design. But today, you can delegate that task to an agentic system. It will research the role, the company, even the market, and return with a contextualised simulation in under an hour. It’s not magic – it’s the compounding effect of rapid cycles, rising literacy, and sharper expectations.

As ever when we’re grappling with newness and struggling to find the shared language, there are acronyms and new terminology flying around – MCPs, agentic systems, boss GPTs. You’ll hear about them. But don’t get stuck on the terminology. This isn’t a glossary moment. It’s a design moment.

You don’t need to know how all the plumbing works. But you do need to think about what you’d do differently if your tools could suddenly see, decide and act. You need to understand the architecture – not the code, but the pattern. Because that’s what’s changing.

We’re not just building tools. We’re building systems that interact with other systems. That’s the big shift. This phase of AI isn’t just technical — it’s architectural. It’s not about knowing every tool, but seeing the shape of what’s coming next.

So, what should leaders do now?

• Upgrade your literacy. Start using the tools more. Explore platforms like Cursor and Manus.

• Review your assumptions. If something feels impossible or too manual today, ask: will it still feel that way in two months?

• Design for change. Don’t build static systems. Build in learning loops. Make adaptability the norm.

This isn’t a time to wait and see. It’s a time to experiment. The next wave will make more things possible. Your job is to be ready to recognise them when they arrive.

And then? Let go of the old way. Quickly.