The Stack: My Approach to Intelligence-Native Architecture

A companion to The Intelligence Bottleneck Is Ending manifesto

When I published my manifesto about the end of the intelligence bottleneck, I deliberately stopped at the vision. The constraint that shaped all human organisation is dissolving. Intelligence is becoming abundant. The companies being built today will operate on fundamentally different principles.

But the vision is only a starting point. How do we actually get there?

Nobody Has the Blueprint Yet

Everyone's talking about AI transformation. Few are building it systematically. Most are bolting AI onto existing structures—the equivalent of replacing steam engines with electric motors but keeping the same belt-driven factory layout.

The real opportunity isn't in making old processes faster. It's in designing entirely new ways value gets created when intelligence is no longer scarce.

Nobody has the complete blueprint for intelligence-native companies. We're discovering it by building it. This post is my contribution to that discovery—a systematic approach I'm using to transform companies from scarcity-era structures to intelligence-native operations.

The Stack: Three Layers, One System

The Stack is the operating system I'm building for intelligence-native companies—three interconnected layers that transform how businesses create and capture value when one person can orchestrate the work of hundreds.

I'm actively working across all three layers —I'm discovering what works by working with real problems. The principles are clear; the implementation is evolving. You can start building intelligence-native operations now with these foundations, even as the full stack matures.

The Unbundling Insight

Before diving into the layers, here's the core mechanism that makes everything work:

When intelligence was scarce, we bundled high-value judgment with low-value tasks into "roles." A QA engineer reviews critical security risks AND checks for typos. A designer creates breakthrough concepts AND resizes assets. A strategist defines market positioning AND formats PowerPoint decks.

With abundant intelligence, we can unbundle. Isolate the vital 20% where human judgment and taste creates exponential value. Scale the other 80% through intelligent processes.

One architect can now orchestrate what used to require a hundred executors. Not through automation, but through a fundamentally different division of labor between human judgment and machine intelligence.

This is the leverage mechanism The Stack enables.

Layer 1: Business Foundation

I start with the Unit of Excellence—the atomic unit where value lives. For a 3PL, it's a job delivered. For a management consultancy, it might be a transformation completed. For a wine producer, it's a wine carton sold. This Unit of Excellence frames the north star eg Profit per Case Sold, the shared metric both employees and agents optimize toward.

The approach applies 80:20 thinking systematically: find the vital 20% that drives 80% of value, instrument it obsessively, improve it continuously.

This isn't a new experiment. It's 20+ years of proven business excellence principles, redesigned for intelligence abundance with a tight focus on high leverage outcomes. The same rigorous measurement and continuous improvement that works for human-scale operations, adapted for when one employee can orchestrate the work of many.

Whether you build your own approach or use established frameworks like OKRs or EOS—what matters is having a foundation that can handle radical leverage, rapid learning, and distributed intelligence.

Layer 2: Intelligence Infrastructure

The second layer is orchestration—creating intelligent processes where outcomes are delivered through durable workflows. Each process knows exactly what success looks like for its outcome. Every action is observable, every decision traceable, every pattern learnable.

This isn't about agent swarms pretending to be human teams. It's about simple patterns that compose into sophisticated behavior:

Atomic knowledge blocks that continue to update
Processes that own their context and evolve their execution
Hierarchies of processes holding context at the right altitude, spotting and acting on patterns lower level processes can't see

No complex choreography. Just intelligent primitives that scale.

Layer 3: Work Transformation

When it comes to GenAI, I've stopped thinking in roles and started thinking in processes. A code review isn't one monolithic task—it's security analysis, performance regression detection, architecture validation, and pattern recognition. Unbundle them. Scale what's repeatable, preserve what requires judgment.

Build process hierarchies where intelligence works at multiple levels. Small teams achieving 100x leverage because they're orchestrating intelligence, not managing people.

Intelligent Processes, Not Artificial People

What I'm discovering: the breakthrough isn't in making AI agents that act like humans. It's in creating processes that think—and remember.

Resolve Ticket doesn't just route to agents—it knows this customer's history, their communication style, their value to the business. It learns that enterprise clients need detailed technical responses while SMBs want quick fixes.

Collect Payment doesn't just send reminders—it remembers that this customer always pays after their monthly board meeting, while another responds better to gentle nudges than stern warnings.

Win Customer doesn't just track pipeline stages—it recognizes that this prospect values ROI discussions while another cares about implementation speed, and adapts its approach accordingly.

Every process becomes a learning system with memory. Each interaction enriches its understanding not just of what works generally, but what works for whom. The organisation evolves continuously, not through change management initiatives but through accumulating intelligence about every relationship.

The Individual as Leverage Point

Here's what most people miss: this isn't about replacing humans. It's about radical leverage of human capability.

The same AI that creates expensive complexity in a confused organization creates exponential leverage in an aligned one. The difference isn't the technology—it's the foundation you build it on.

The companies building on approaches like The Stack today aren't competing on features or efficiency. They're operating on different physics—where traditional constraints don't apply and new ones (alignment, taste, judgment) become everything.

Building the Stack

The intelligence bottleneck that shaped every business for all of human history is ending. The question isn't whether to adapt, but how systematically you'll approach the transformation.

I'm building this operating system across three layers:

Business Foundation — Define your Unit of Excellence. Instrument it. Build the feedback loops that let both humans and agents optimize toward the same ends.
Intelligence Infrastructure — Durable, intelligent workflows, processes that learn and remember.
Work Transformation — Unbundle roles into processes. Scale what's repeatable, preserve what requires judgment.

The complete stack is still emerging. We're discovering what intelligence-native means by building it.

Building Intelligence-Native?

I work with companies across all three layers of the stack—business foundation, intelligence infrastructure, and work transformation. If you're ready to architect for abundance rather than scarcity, let's talk.

Part of The Intelligence Shift. Subscribe for weekly insights on intelligence-native transformation: Join the newsletter