Welcome to the Intelligence Era

What is actually being built, why almost nobody is explaining it, and why understanding it is the most important thing you can do right now

AI. It is in every meeting, every pitch deck, every earnings call, every product announcement. Every vendor you already pay for has added an “AI powered” badge to something that worked the same way last year. Every conference has an AI track. Every board wants an AI strategy. The noise is relentless.

I get it. I am in it. I speak to executives, founders, and operators every single day who are trying to make sense of what is happening, and the same frustration comes up in almost every conversation.

They know something fundamental has shifted. They can see the headlines. They have tried the tools. But when they sit down at their desk in the morning, the work feels the same. Eleven tabs. Copy. Paste. Reformat. Send. The tools are louder about being intelligent. The actual work has not changed.

And that gap between what they are being told is happening and what they are actually experiencing is making people feel like they are either missing something obvious or being sold something that does not exist.

They are not missing anything. And it is not vapourware. The gap is real, and it has a very specific cause. Understanding that cause is the most important thing I can share with you right now, because once you see it, every decision you make about technology, about your team, about your business, starts to look different.

The natural reaction to this kind of uncertainty is to slow down. Run a pilot. Form a committee. Commission a strategy review. Wait for clarity before committing.

That instinct is understandable. It is also the single most dangerous thing you can do right now.

The bicycle problem

There is an old observation about riding a bicycle that Nate B. Jones highlighted recently, and it has stuck with me ever since. Go slowly and you wobble. The handlebars feel unstable. Every small input creates an overcorrection. You grip tighter, which makes it worse. It feels like the problem is balance.

It is not. The problem is speed.

Go faster and the bicycle stabilises itself. The physics change. Momentum creates its own equilibrium. The same machine that felt dangerous at walking pace becomes effortless at speed. (If you are not already following Nate’s work on AI strategy, you should be. He is one of the clearest thinkers in the space.)

This is exactly where most organisations sit right now.

They are moving slowly through a transition that punishes hesitation. Every week a new model launches. Every month the competitive landscape shifts. The committee meets again. The pilot gets extended. Nothing ships. And the wobble intensifies, not because the technology is immature but because the pace of decision making is wrong for the pace of change.

The organisations that have found their balance are the ones that committed to velocity. Not recklessness. Velocity with direction. They picked a problem, built something, learned from it, and built the next thing. They are not smarter. They are faster. And the speed itself is what makes the decisions clearer.

Clarity does not come before action. It comes from it.

This piece is for everyone still gripping the handlebars. If the last article explained what happened, this one explains what you are looking at and why seeing it clearly is the first thing that matters.

We are not upgrading. We are replacing the architecture.

Here is the distinction that most people are missing, and once you see it, it changes how you look at everything.

An upgrade improves what exists. You move from one version of software to the next. The logic stays the same. The workflows stay the same. The fundamental relationship between the human and the machine stays the same. Just faster, cleaner, with a few more features.

An architectural transition replaces the logic itself. The relationships change. The workflows change. What was once the product becomes the infrastructure. What was once the infrastructure becomes invisible.

We went through this before. When companies moved from on premise servers to the cloud, they did not just get cheaper storage. The entire model of how software was built, sold, and delivered changed. SaaS emerged. Per seat licensing appeared. Companies that understood the architectural shift built trillion pound businesses. Companies that treated the cloud as “someone else’s server” missed the point entirely.

The same thing is happening now, except the shift is larger.

We are not adding AI to the existing stack. We are replacing the stack’s operating logic. In the classical model, applications contained the intelligence (such as it was) and humans operated them. In the emerging model, intelligence sits above the applications and orchestrates them. The applications still exist. But they become infrastructure: databases and services that an intelligent layer calls upon, rather than interfaces that humans navigate.

This is not a feature upgrade. This is the end of one architectural era and the beginning of another.

And if you do not see the distinction, every decision you make from here will be slightly wrong.

The illusion of intelligent tools

Every major SaaS vendor is now racing to build AI into their product. Salesforce has rebranded its entire AI strategy around Agentforce. HubSpot has AI assistants. Microsoft has Copilot woven through the entire Office suite. ServiceNow, Workday, SAP: everyone is adding an intelligence layer inside their own walls. On the surface, this looks like progress.

Look closer.

Each of these tools can only see what lives inside its own walls. Salesforce’s AI knows your CRM data. It does not know your project timelines, your financial forecasts, your email conversations, your Slack threads, or the recording of last Tuesday’s board meeting where the CEO changed strategic direction.

HubSpot’s AI knows your marketing data. It does not know that your biggest client is about to churn because of a support issue logged in Zendesk, discussed in a WhatsApp group, and escalated in a Monday morning standup that nobody documented.

Each tool is building intelligence inside a silo. And an intelligent silo is, in some ways, worse than a dumb one. A dumb silo just stores data. You know it cannot think, so you do the thinking yourself. An intelligent silo gives you the impression of comprehension without the reality. It offers recommendations based on a fraction of what matters and presents them with the confidence of a system that believes it has seen everything.

Confidence without context is not intelligence. It is hallucination with a dashboard.

The problem is not the models. The models are extraordinary. The problem is that context does not live inside any single application. It lives across all of them, in the spaces between them, in the conversations and decisions and institutional knowledge that no individual tool was ever designed to hold.

This is why bolting AI onto existing tools, however sophisticated the AI, does not get you where you need to go. It solves the wrong problem at the wrong layer.

What intelligence architecture actually looks like

So what does the right architecture look like? Not in theory. Not in a consultant’s slide deck. In the real world, where things have to work.

Think of it as a stack with five layers. Each one serves a different purpose. Each one is necessary. And the value concentrates in a place that will surprise almost everyone still thinking about AI in terms of which model to use.

The foundation layer is where the large language models live. Claude, GPT, Gemini, Llama, Mistral. These are the engines. They are extraordinary and they are rapidly being commoditised. A year ago, model selection felt like the most consequential decision in AI. Today, the frontier models are converging in capability and the smartest builders are treating them as interchangeable utilities, routing different tasks to different models based on cost, speed, and capability. This layer matters. But it is not where competitive advantage lives.

The context layer is everything the business knows, made accessible to intelligence. Documents, data, processes, institutional memory, customer history, strategic direction, the recording of last week’s all hands. Most of what makes a business distinctive lives in this layer, and most of it is currently trapped in silos, in people’s heads, in email threads, in meeting recordings that nobody watches twice. Building this layer is the hardest and most valuable work in intelligence architecture. It is not a technology problem. It is an organisational problem. The companies that solve it will have an advantage that compounds over time, because every day their intelligence gets smarter while their competitors’ AI remains generic.

The orchestration layer sits at the centre. It routes requests to the right model. It manages memory. It coordinates agents. It decides what context to retrieve and when. It handles the conversation between multiple AI systems working together. If the context layer is the knowledge, the orchestration layer is the judgment about how to use it. This is the layer where open standards are emerging fastest, and it is the layer where the biggest strategic battles in technology are currently being fought. More on that in a moment.

The action layer is where agents live. Agents that can draft communications, manage workflows, analyse data, make recommendations, and increasingly take action on your behalf. Not chatbots. Systems that do things. The distinction matters. A chatbot waits for you to ask a question. An agent monitors, reasons, and acts within boundaries you define. This is where the “intelligent silo” problem gets solved: a well designed agent does not live inside one application. It sits above all of them, pulling context from wherever it lives and taking action across whatever tools are needed.

The interface layer is now the thinnest and most disposable part of the entire stack. It used to be everything. The entire SaaS era was built on the premise that the interface was the product. Now the interface is just where you happen to talk to the intelligence. It might be a web app. It might be WhatsApp. It might be your email client. It might be voice. The interface will be vibe coded, rebuilt, and replaced more often than any other component, because it no longer carries the value.

The interface is the window. Not the building.

Here is the inversion that will reshape every technology decision you make: in the classical era, the interface was the product and the data was the byproduct. In the intelligence era, the context is the product and the interface is disposable. If your organisation is still spending the majority of its technology budget on interfaces, you are investing in the one layer that is about to matter least.

The race underneath the surface

If you understand this stack, you can see what the major technology companies are actually fighting over. And it is not what most people think.

Anthropic, OpenAI, and Google are not competing primarily on model quality. They are competing to own the orchestration layer. The memory. The tool ecosystem. The protocols. The relationship between intelligence and the world it acts upon.

MCP, Anthropic’s Model Context Protocol, became the industry standard for agent to tool communication in under twelve months. It launched in November 2024. By mid 2025, OpenAI had adopted it. Google had adopted it. Microsoft had adopted it. Thousands of developers had built on it. Then in December 2025, Anthropic donated the entire protocol to the Linux Foundation through the newly formed Agentic AI Foundation. Shortly after, Anthropic launched Agent Skills as a separate open standard, which OpenAI integrated almost immediately.

Ask yourself: why would a company give away something this valuable?

Because controlling the standard is more powerful than controlling the product. The company that defines how agents connect, remember, and act is not building a feature. It is building the operating system for the intelligence era. And operating systems tend to be permanent.

This is the most consequential infrastructure battle since the early web. And most people are watching from the wrong angle, focused on which model scored highest on a benchmark rather than who is quietly building the plumbing that everything else will run on.

But here is the part that should excite you rather than intimidate you. Open standards mean open opportunity. MCP is open source. The protocols are documented. The tools to build on top of this infrastructure are available to anyone with the curiosity and determination to learn them.

You can already see the evidence. In January 2026, an open source project called OpenClaw became the fastest growing repository in GitHub history: 157,000 stars in sixty days, eighteen times faster than Kubernetes. Built by a single Austrian developer, it gave anyone the ability to run a powerful AI agent across their entire tech stack, locally, privately, connecting to whatever models and tools they chose. It brought all the dangers and rough edges you would expect from something moving that fast. But the community is solving those problems the same way the open source community has always solved them: in the open, together, iteratively.

This matters because it reveals something fundamental about where the value will live. Most of the internet’s core infrastructure was built on open source. Linux. Apache. Kubernetes. The pattern repeats. The orchestration layer of the intelligence era will follow the same path. Open protocols. Open standards. Private companies building proprietary value on top of open foundations. The models will be rented. The context will be owned. And the orchestration layer, the one that connects everything, will be open because it has to be. No single company can own the connective tissue of intelligence without breaking the system.

The playing field has not been this level since the early days of the web.

Which brings us to a question that every organisation, regardless of size or sector, needs to start asking: what is our intelligence architecture?

Not “what AI tools are we using?” That is the wrong question. The right question is: how does our proprietary knowledge, our institutional context, our data, connect to the intelligence layer? Every company needs to start thinking of itself as an API. Not a technology company. But a company that understands how to make its own knowledge accessible, structured, and ready for intelligence systems to use. The organisations that figure this out will compound their advantage. The ones that do not will find their competitors’ AI getting smarter every day while theirs remains generic, because the models are the same for everyone. The context is what makes the difference.

This requires people inside the business who understand the architecture. Not an army of software engineers. But people who can see how the pieces connect, who can bridge the gap between what the business knows and what the intelligence layer needs. People who can set their organisation up for what comes next.

Most companies do not have these people yet. The ones that find them first will have a head start that will be very difficult to close.

The generalist’s moment

For decades, the technology industry has rewarded specialists. Deep expertise in one language, one framework, one platform. The person who knew Salesforce inside out was more employable than the person who understood a bit of everything. The specialist was safe. The generalist was told to pick a lane.

That world is inverting. Fast.

When 41% of all code is now AI generated and 92% of developers use AI tools daily, the ability to write syntax from scratch is no longer the scarce resource. What is scarce is the ability to see across systems, to understand how business context connects to technical possibility, to define what good looks like and orchestrate the intelligence that delivers it.

This is generalist territory. And it is the biggest opportunity of a lifetime for people who have always thought broadly rather than deeply.

The people who will thrive in the intelligence era are not prompt engineers. That concept is already dated: it reduces the work to a parlour trick of word selection when the real skill is architectural thinking. They are not traditional software engineers in the narrow sense either, though understanding how systems work remains enormously valuable.

The new role looks something like this: someone who can articulate a business problem clearly, who understands enough about how intelligence systems work to know what is possible and what is not, who can define the boundaries within which an agent should operate, who can provide the proprietary context that no foundation model has, and who can evaluate whether the output actually solves the problem it was meant to solve.

Some people are calling them intelligence engineers. Others call them forward deployed engineers or AI architects. The title does not matter. What matters is the shape of the thinking: broad, contextual, outcome oriented, and comfortable moving between the business problem and the technical solution without getting stuck in either.

If you have spent your career being told you are too broad, too curious, too interested in too many things: the world just reorganised itself around the way you think.

The specialist era rewarded depth. The intelligence era rewards range. And the people who spent years feeling like they did not quite fit are about to discover they were training for a job that did not exist yet.

More builders, not fewer

Here is where the story turns towards something thrilling, and it is worth staying here for a moment because the dominant narrative around AI and work is fear. Fear of displacement. Fear of irrelevance. Fear that the machines will take the jobs and leave nothing behind.

The evidence points somewhere entirely different.

We are going to have more software, more code, more products, more services, and more builders than at any point in human history. Not fewer. More. Dramatically, almost incomprehensibly, more.

Vibe coding went from a joke by Andrej Karpathy to Collins Dictionary’s word of the year in under twelve months. People with no formal engineering background are building working products over a weekend for the cost of an API subscription. Small businesses that could never afford custom software are creating tools tailored to their specific needs. Entrepreneurs who previously had to raise funding just to build a prototype can now validate ideas in hours.

This is not the death of software engineering. Software engineers who understand code, architecture, and systems thinking are more valuable than ever because the volume of software being created is exploding and someone has to ensure it works, scales, and does not collapse under its own weight. What is changing is who gets to participate in building. The barrier did not just lower. It collapsed. And the people walking through are not just engineers. They are domain experts, business operators, creative thinkers, and problem solvers who never had the tools to turn their ideas into working systems before.

Think about what happened when the printing press appeared. It did not eliminate writers. It created an explosion of publishing, literacy, and entirely new forms of writing that had never existed. The people who feared it were the scribes who thought their monopoly on the written word was the point.

It was never the point. The point was the ideas. It was always the ideas.

The same dynamic is playing out now, at a pace that makes the printing press look leisurely.

Why understanding this is not passive

If you have read this far, you might be tempted to treat what you have just learned as background knowledge. Interesting. Filed away. Something to mention in the next leadership meeting.

That would be a mistake.

Understanding the architecture of intelligence is not an intellectual exercise. It is the single most consequential shift in perspective you can make right now, because it changes every decision that flows from it.

Once you see the five layers, you stop evaluating AI tools by their features and start evaluating them by which layer they occupy. You realise that your CRM vendor’s “AI assistant” is solving the wrong problem at the wrong layer, and that the real question is not which tools to buy but how to make your organisation’s knowledge accessible to an intelligence layer that does not exist yet.

Once you understand that context is the competitive moat, you start asking different questions about your own organisation. Where does your institutional knowledge actually live? How much of it is trapped in people’s heads, buried in email threads, locked inside applications that do not talk to each other? Most organisations have not even mapped what they have, let alone made it accessible. The ones that start mapping it now will have an advantage that compounds every single day.

Once you see the orchestration layer, you understand why this is not something you solve by hiring a single AI vendor or appointing a Head of AI. It is architectural. It touches everything. And it requires people who can see across the entire landscape of your business and your technology and understand how the pieces connect.

This is where most organisations hit the wall. Not because the technology is too complex. Because they do not have people around them who understand how to think about it.

The companies that do get it are already building. One example I have seen up close is Flourish, a platform that helps global brands deliver compliant marketing by unifying regulatory knowledge, brand evidence, and AI agents into a single intelligence layer. Full disclosure: I advise them. But the reason I chose to is because they are a textbook case of a company that understood the architectural shift before most of the market did. They did not bolt AI onto an existing compliance workflow. They built the context layer first, then the orchestration, then the agents. That is the pattern. And these are the kinds of products and companies that people should be studying, building, or partnering with.

The transition to intelligence architecture is not a solo endeavour. It requires a new kind of conversation: one that bridges business strategy, technical architecture, and organisational design all at once. If you do not have people in your world who can hold that conversation, finding them is the single most valuable thing you can do next.

Not next quarter. Now. Because every week that passes, the organisations that do have those people are compounding their understanding, and the gap is getting wider.

The window

We taught the machines everything. They woke up. And now we are standing inside an architectural transition that most people will only recognise in hindsight.

The intelligence era is not arriving. It is here. The protocols are being written. The standards are being set. The early movers are compounding their understanding with every week that passes. And understanding is not a passive state. It is the prerequisite for every decision, every investment, every hire, every conversation that will determine whether you lead this transition or get dragged along by it.

This is not a warning. It is an invitation.

The intelligence era does not eliminate human contribution. It elevates it. It takes the drudgery, the copy and paste, the manual translation between systems, and it automates it. What remains is the work that actually requires judgment, creativity, and contextual understanding.

The work that matters. The work that was always the point.

If you are an executive, this is the moment to stop treating AI as a feature request and start treating it as an architectural decision. If you are a builder, this is the greatest playground you will ever have. If you are a generalist who spent years being told to specialise, this is your vindication.

The classical era taught us how to organise information. The intelligence era will teach us what to do with it.

The question was never whether the machines would learn. They learned plenty.

The question is whether we understand what they have made possible. And whether we are ready to act on it.

---

This is Part Two of a series. Part One, “We Taught the Machines Everything. Now What?” explored how four decades of human work created the foundation for the intelligence era. Part Three will go deeper into the context layer: why it is the true competitive moat, how the organisations building it today are creating advantages that will be nearly impossible to replicate, and what the first practical steps look like when you are ready to build.

Craig Hepburn is Co-Founder & CEO of RAIN Ventures, an AI venture studio building intelligent systems and “useful intelligence” for businesses. He writes about the strategic implications of AI at the intersection of technology and business transformation.