Principle 1
The 5 Profit Drivers
These are the short-term metrics that move when you push workflows up the maturity ladder. The 0-12 month scorecard. If you move meaningful work from Level 1 to Level 5 with production discipline (safe, measurable, governed), these five drivers move up and to the right.
Unit Economics
Fewer write-offs, less rework, more standardization. As workflows mature, the cost to produce each unit of output drops because the work is repeatable and quality-controlled. At Level 1, every deliverable is a custom job. At Level 4, the agent runs the workflow and the unit cost approaches zero.
Throughput and Capacity
More output per person. Fewer bottlenecks. At Level 1, capacity is limited by headcount and heroics. At Level 3, AI accelerates steps inside the SOP. At Level 5, agent swarms operate across functions simultaneously. Capacity becomes a function of how many workflows you have matured, not how many people you employ.
Time-to-Value
Shorter cycle times. Faster responsiveness. The coordination tax (handoffs between people, functions, and external vendors) shrinks as workflows mature. Work that took weeks takes minutes. The gap between "idea" and "first usable iteration" collapses.
Reliability and Risk
Fewer defects. More traceability. Consistent QA. At Level 1, quality depends on whoever does the work that day. At Level 3+, the workflow enforces quality standards. Embedded QA prompts, checklists, and definitions of done make the output predictable.
Cash Conversion
Less WIP aging. Faster completion leads to faster billing and collections. When cycle times compress, the time between starting work and getting paid for it shrinks. Cash flow becomes more predictable because delivery is more predictable.
These are structural outcomes. If you standardize a workflow and automate it with production discipline, these five things happen mechanically. The performance outcome is structural.
Principle 2
The 5 Enterprise Value Drivers
These are the long-term compounders. The 1-10 year metrics. What makes a company worth more over time. While the profit drivers show up in the next quarter, these show up when someone evaluates the company for acquisition, investment, or strategic positioning.
Learning-Rate Compounding
Improvements become reusable assets. Every time you run a workflow, you learn something. That learning gets codified back into the system. The system improves. The next run is better. Over time, you are not just doing work. You are improving the factory that does work. Work produces artifacts. Artifacts produce insights. Insights produce system improvements. System improvements make the next round of work better.
Market Capture Flywheel
Better experience leads to retention, referrals, and higher win-rate. When your delivery is faster, more reliable, and more consistent, clients stay. They refer others. Your close rate goes up because the proof of capability is in the work itself. This compounds over years.
Strategic Optionality
New offers and tiers ship faster because delivery is standardized. When your workflows are mature, you can create new products by combining existing capabilities. You can enter new markets faster because the production infrastructure already exists. Standardized delivery gives you the option to move in directions your competitors cannot because they would have to build from scratch.
Talent System Advantage
Onboarding and performance become system-driven, not hero-driven. New people get productive faster because the workflows are documented and the agents do the heavy lifting. You are less dependent on finding and retaining individual stars because the system carries the institutional knowledge. Key-person risk drops.
Expansion and Consolidation Power
Replicable operating model across teams, locations, and acquisitions. Once your operating model works, you can replicate it. Open a new office, onboard an acquisition, expand into a new geography. The playbook travels because it is in the system, not in people's heads.
Profit drivers are about doing the work better today. Enterprise value drivers are about building a company that is worth more over time. The maturity ladder drives both simultaneously, but on different time horizons.
Principle 3
The 5 Levels of Workflow Maturity
Every company is running a portfolio of workflows. Every workflow sits somewhere on a five-level maturity ladder. The core argument: if you move meaningful work from Level 1 to Level 5 with production discipline, you push all profit drivers and enterprise value drivers up and to the right.
Level 1: No SOP, Manual Work
Tribal knowledge. Quality varies by who does the work. Rework, bottlenecks, scale by hiring or heroics. Key-person risk blocks enterprise value. No compounding loop because nothing is written down.
Level 2: Manual Work with SOP
The breakthrough. Work becomes describable. SOPs define "what good looks like," create a repeatable path for average performers, and turn workflows into assets that can be improved. This is the biggest move on the ladder because without Level 2, you do not have clean inputs for Levels 3-5. AI makes SOP creation cheap. What used to be a multi-week burden becomes a rapid sprint.
Level 3: AI Workflows
Humans still run the work, but AI accelerates steps inside a repeatable SOP. Faster drafting and analysis, standardized outputs, embedded QA prompts. Cycle time drops, capacity rises, and rework falls because the workflow is now repeatable and augmented.
Level 4: AI Agents
Agents execute defined workflows end-to-end or in large sections with clear boundaries. Humans supervise and handle exceptions. At Level 4, AI capability is not the limiting factor. The limiting factor is missing or messy context, unclear definitions of quality, lack of auditability, and no improvement loop. Every agent must be grounded in production primitives: process, examples, context manifests, QC.
Level 5: AI Agent Swarm
Multiple agents collaborate across workflows and functions under orchestration, shared standards, continuous monitoring, and continuous improvement. Knowledge becomes portable. Onboarding accelerates. New offerings ship faster. Operations become replicable. Level 5 is system-level coordination plus system-level learning.
Why this is a law
What increases across the ladder is standardization plus leverage plus repeatability. Those are the structural causes of higher margins, more throughput, faster cycle time, lower defect rates, faster cash conversion, and compounding capability over time.
The practical wedge
The biggest move is Level 1 to Level 2 because AI collapses the cost and time required to document work. Most companies avoid SOPs because they feel like a tax. AI makes SOP creation cheap enough to become an obvious first step. From there, moving up the ladder becomes less about "figuring out AI" and more about disciplined execution. You are taking what is now written down and turning it into repeatable machine-assisted production.
The ladder is the operating model. Once you commit to moving work up levels, the performance outcome is structural.
Principle 4
The 5 Stages of AI Development
This framework maps out five stages over roughly a ten-year arc that started in November 2022 when ChatGPT launched. Once you internalize this arc, it changes how you plan, how you invest, and how you think about what is coming next.
Stage 1: Chatbots (Started November 2022)
You type something, the AI responds. Back and forth conversation. Incredibly useful for many tasks, but fundamentally reactive. Millions of people got their first taste of AI in this phase.
Stage 2: Reasoning
Chain of thought reasoning models. The AI shows its work, reasons through problems step by step. Much more capable than chatbots. Things that took multiple prompts and back and forth could now happen in a single interaction.
Stage 3: Agents (Current stage)
Agents take goals, break them down, execute tasks, use tools, and solve problems without constant human intervention. The shift from "AI responds to me" to "AI works for me" changes everything about how you can operate a business.
Stage 4: Innovators
AI discovers new solutions. Drug discovery. Cures for cancer. But also inventing new business products and services. The AI does not just execute what you tell it. It invents new ways to do things.
Stage 5: Organizations
AI runs entire organizations. AI as the primary agent making decisions and executing strategy.
The trajectory matters more than any snapshot of where we are today. If you only look at what AI can do right now, you will underinvest and fall behind. If you understand the arc, you can position yourself ahead of it.
Now that agents actually work well, capabilities will not improve linearly anymore. They will accelerate. Each improvement makes the next one easier and faster. This is the hard takeoff.
Principle 5
The 5 Levels of Knowledge Work
To understand what AI agents will eventually do for you, you have to see the architecture of knowledge work. Break it into its component parts. Once you can see the pieces clearly, you can see where AI fits in, and where the compression happens.
Level 1: Objectives / Goals
Where you want to be. What success looks like.
Level 2: Projects
The initiatives you run to meet those objectives. A project has a clear scope, timeline, and deliverables.
Level 3: Deliverables
The specific outputs from each project. A report. A strategy document. A feature. A design. Something tangible.
Level 4: Tasks
The work required to produce each deliverable. Research, writing, analysis, testing. Discrete units of work.
Level 5: Actions
The smallest units. Opening a file. Making a phone call. Writing a sentence. The atomic unit of effort.
Plan Down, Do Up
When you plan, you work your way down the ladder. Start with objectives, break into projects, break into deliverables, break into tasks, break into actions. When you execute, you work your way back up. Complete actions, which roll up into tasks, which roll up into deliverables, which complete projects, which move you toward objectives. Continuous improvement is about running this loop as fast as you can.
AI Compression Across the Stages
In the chatbot phase, planning from objectives down to actions took about fifty well-formed prompts and hours of work. With reasoning models, one prompt with about two minutes of thinking time produced the same result. In the agent phase, you give an agent a goal, set it loose, and an entire project gets done in a single run. As we progress through the five stages of AI development, both planning and doing get compressed. The loop runs faster. You can run it more times in the same amount of calendar time.
Principle 6
The 5 Orders of Time Magnitude
Time is changing. Not just that things are getting faster. The fundamental unit of time in which economically valuable work gets done is shifting. There are two dimensions moving simultaneously.
Dimension One: Task Time Compresses Down
Things that used to take weeks now take minutes. Chatbots compressed weeks down to days. Reasoning compressed days down to hours. Agents are compressing hours down to minutes. But be precise about what counts: an economically valuable task is one where the actions complete tasks that complete deliverables that complete projects that actually meet your objectives. The AI voice agent making thousands of calls is not producing economically valuable work if none of those calls book a qualified lead.
Dimension Two: Autonomous Work Time Expands Up
In the chatbot stage, autonomous work time was seconds. You type, it responds, you type again. In the reasoning stage, it grew to minutes. In the agents stage, it is hours or even days. You give an agent a goal, set it loose, and it works autonomously while you do something else entirely. This enables real multitasking. Launch multiple agents simultaneously. Review output from earlier runs while new runs are in progress. Your job shifts from doing the work to orchestrating the work.
The CEO Time Shift: As a CEO with deep knowledge of your entire business, the bottleneck was never knowledge or capability. You knew what needed to be done. You often knew exactly how to do it. The bottleneck was always time. There simply were not enough hours in the day to do everything your brain was capable of directing. AI changes that equation. You now have time. That changes everything.
Principle 7
The Information Factory Stack
You are not just using AI as a tool. You are building a factory. An information factory. The planning and doing from the knowledge work principle get compressed so you can find economically valuable work faster. Then the time dynamics kick in: the time to do that work shrinks while the AI's ability to work autonomously expands. Put all of that together and you have a system that produces information at scale.
Layer 1: Humans
Always the foundation. Humans set the goal for the AI.
Layer 2: LLMs
Only a handful of frontier models exist: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), plus a few minor players. Every other AI tool on the planet is built on top of these.
Layer 3: Workflows
Work instructions for what an agent does. Sometimes predetermined, sometimes not. A series of steps an agent goes through to get work done.
Layer 4: Data
Context the agent needs access to while working.
Layer 5: Tools
What the agent uses while doing its work. APIs, integrations, systems.
The Industrial Revolution Parallel
Before the industrial revolution, the economy was driven by artisans. Everything was local. Everything was handmade. Every craftsperson had spent years honing their skill. Then factories arrived. You could take someone with no training, put them in front of a machine, give them work instructions, give them raw materials, and make sure they had the right tools. And that person could produce output. The person off the street is the human. The machine is the LLM. The work instructions are your workflows. The raw materials are your data. The tools are the tools.
Once the output coming out of the factory reaches equal or better quality than what the craftsman used to make by hand, the economics become inescapable. You are compelled to build an information factory or you simply cannot compete.
The leverage is in the layers you build on top: your workflows, your data, your tools.
Principle 8
The 5 Stages of Technology Diffusion
Just because something is possible does not mean it will instantaneously spread across the economy. Things take time to diffuse. This idea stretches back to the 1950s, and understanding it gives you a massive strategic advantage.
Stage 1: Knowledge
Someone needs to know about it. That takes time.
Stage 2: Persuasion
They need to be persuaded it is worth the effort to try. Even I take time before trying things I just heard about, because there are a million other things to do.
Stage 3: Decision
Run a full-blown pilot. Pull decent resources. There is an opportunity cost.
Stage 4: Implementation
Actually implement and see whether it works.
Stage 5: Confirmation
Confirm it works and decide if it is something you are going to do over the long run.
The Critical Twist for AI
AI gets exponentially better over time. It is a new technology every time there is a step change improvement. And those step changes are happening faster and faster. The rate of change is now compressing into weekly cycles. But just because you know about something does not mean the rest of the economy does. You are in a very small percentage of the population.
Diffusion gives you a runway of months. Even in a business at the frontier, it takes time for new capabilities to fully diffuse through the organization. Then it takes several more months for those capabilities to diffuse to the rest of the economy. If you can see where everything is going, you do not need to frantically try every new thing the moment it appears. You can be thoughtful and methodical and still be months ahead of everyone else.
Principle 9
The 5 Stages of AI Acceptance
Every time the technology makes a significant jump, these stages reset. You go through them again. Your team goes through them again.
Stage 1: Indifference
The new thing has not touched your work yet. It is someone else's problem.
Stage 2: Fear
You see AI do something at equal or higher quality than you could, in a massively compressed time frame. You get scared. This thing can do what I do.
Stage 3: Letting Go
This is where most people get stuck. You can stay in fear for a very long time, or you can let go of the part of your ego that attached itself to "I am valuable because I can do this thing." The people who move through it come out with more confidence and capability because they are using AI to amplify what they can do rather than competing against it.
Stage 4: Confidence
You start using AI to do that thing you used to do yourself. You are getting so much more done. Confidence builds as your output multiplies.
Stage 5: Agency
The ability to go off and solve problems you previously could not solve, because you now have the bandwidth. This is where enterprise value creation accelerates.
A Reframe on Higher Order Work
AI does not free you to do some predefined category of elevated work. AI frees you from whatever it can do better than you can. You then move on to the next bottleneck, whatever that happens to be. Sometimes that is strategic thinking. Sometimes it is having difficult conversations with your team. Sometimes it is just shipping faster. AI gives you the time to work on whatever the actual constraint is, and the agency to solve problems you previously had no business trying to tackle.
The Leadership Dimension
With good leadership, this transition can happen remarkably quickly. People see you modeling the behavior. They see you using AI with confidence. They see the results. And they follow. With bad leadership, it happens very slowly, with a ton of resistance. And every time AI gets better, you personally have to deal with these stages again. You have your own ego deaths to work through. Then you have to help your team work through theirs.
Principle 10
The 5 Levels of AI Insourcing
For decades, professional services firms charged premium prices because they had concentrated subject matter expertise that was very difficult to replicate internally. That bottleneck is gone. Post-AI tipping point, AI with a good workflow can produce SME-level work.
Level 1: No SOP (External) / $100
Work done by external professional services. The baseline cost.
Level 2: SOP (Internal) / $33
Bring work in-house. Professional services firms sell at roughly 3x cost. Immediate savings of $67.
Level 3: SME + AI / $3.33
Subject matter expert using AI. At least 10x more efficient. No decrease in quality, often increasing. You can skip Level 2 entirely and go straight here.
Level 4: Non-SME + AI / $0.33
Fine-tune the workflow so you do not need an SME to review the output. Hand it to a virtual assistant at 10x less than a Western worker. Same work. Thirty-three cents.
Level 5: Agent / $0.03
No human in the loop. An AI agent runs the workflow autonomously. Three cents.
The insourcing test is simple: if speed can be increased by an order of magnitude at equal or greater quality, you bring it in. Once you get down to "this can be done by an agent," the decision is not even a real decision anymore. When you start to view all spending inside your company through this math, the whole picture simplifies.
Principle 11
The 5 AI Insourcing Powers
Cost is the entry point for insourcing. It is the most immediately compelling argument you can make to a CFO. But it is the least powerful of the five over the long term. The core rule: we do not want human beings doing work that can be done at the same or greater quality an order of magnitude faster.
Power 1: Cost (Task Time)
You compress how long a single unit of work takes. The work gets done 100 to 1,000 times faster. Massive cost reduction. The CFO metric. The entry point. But the least powerful over time.
Power 2: Velocity (Cycle Time)
Considerably more powerful than cost. The coordination tax is everywhere. Every handoff adds cycle time. An idea used to go from me to creative director to developer and back over multiple weeks just to get to the first moment of real clarity. Now I go through several iterations in ten minutes. Cost is the CFO metric. Velocity is the CEO metric.
Power 3: Control
When you do the work internally with your own SOPs and AI, you control the valuable IP that creates enterprise value. You own the SOP. You know the steps. You control both the current state and the future state. If that IP lives in external heads, that is a massive liability for enterprise value.
Power 4: Elasticity
Once you have people with high AI agency, domain expertise stops being the constraint. You can point your entire organization at the most valuable bottlenecks regardless of background. Someone with zero marketing expertise solved a highly technical marketing problem using AI. Her domain expertise was irrelevant. Her agency was everything.
Power 5: Learning
The one that compounds. Every frontier learning win is a competitive advantage. Each time you solve one constraint, the next one reveals itself, and you are already in motion. Competitors look at what you are doing and think, "How the hell are they doing that?" From the inside, it is obvious. Compounding learning, one win stacking on top of the last.
Principle 12
The 5 AI Agency Pillars
What does the most valuable employee in the AI age look like? There are five skills. People either have them or they do not. They can be taught, but they tend to be innate in some people. For most of business history, these skills were not particularly valued. Now they are extremely valuable.
Pillar 1: Systems Thinking
Once you are unlocked from needing to be a subject matter expert in every domain, you need to think in systems. How does work in one area impact another? Learnings from marketing can be applied to operations. Insights from customer support can reshape product development. Someone has to see those connections and translate them across the business.
Pillar 2: Enterprise Value Focus
Where people point their energy. Now that we can point people at problems that were previously impossible to solve, we want them focused on the problems that compound in value. The ones that build competitive advantage over time.
Pillar 3: Initiative
The ability to go off and solve problems nobody asked you to solve, because you see the value and because experiments can now be run in 20 minutes instead of months. The person who will not wait around to be told what to do is your most valuable player.
Pillar 4: Governance
Risk tolerance combined with the ability to control risk. Take on appropriate risk but know how to govern it so you do not sink the battleship. Help people understand risks, mitigate as many as possible, and know which risks simply cannot be tolerated regardless of upside.
Pillar 5: Adaptive Learning
The ability to learn quickly, but most importantly to focus on learning from what actually happens in the real world. Not theoretical learning. Real-world feedback loops. The person who learns fastest from reality wins.
How They Connect
The 12 as One System
These principles are a coherent system. They become more powerful when you stack them together than any single principle is on its own.
Principles 1 and 2 give you the scorecard: profit drivers move in the short term, enterprise value drivers compound in the long term. Principle 3 gives you the mechanism: the maturity ladder moves both scorecards simultaneously. Principle 4 shows you the trajectory of where AI is heading. Principle 5 deconstructs knowledge work so you can see what agents will do. Principle 6 reveals how time is compressing and expanding simultaneously. Principle 7 shows you the factory you are building. Principle 8 explains why diffusion gives you a runway of months. Principle 9 maps the human acceptance journey you and your team will go through repeatedly. Principle 10 lays out the economics of insourcing. Principle 11 unlocks the five powers that compound beyond cost alone. Principle 12 defines the people who make all of it work.
You see the future because you understand information factories. You move faster than competitors because diffusion gives you months of runway they do not know they are losing. You build competitive advantage because the insourcing powers compound over time. And you attract and develop the right people because you know exactly what an AI-age A-player looks like.