The Mainline Approach
The Factory
Problem
Every organization just got a new engine. Almost none have redesigned the factory.
The Pattern
New Motor, Same Factory
Factories swapped steam engines for electric motors.
No productivity gains for 30 years. They redesigned the motor, not the factory floor. Same layout. Same workflows. Same bottlenecks. Just a different power source.
Companies are swapping in AI tools.
Giving everyone ChatGPT and Claude Code licenses. Same workflows, same processes, same org charts. Individual productivity goes up. Organizational output stays flat.
“Individual AI productivity tools have not made companies 10x more valuable.”
The gains leak out through coordination failures and process debt.
The Evidence
Code Already Figured This Out
The coding world proved that the harness matters more than the model. With data.
0x
improvement from harness alone
Slate V1 Terminal Bench
0M
lines of code, zero human-written
OpenAI internal
42→0%
same model, different harness
Claude Opus benchmark
They stopped at code. What about marketing? Operations? Strategy? Hiring?
The Framework
Five Roles, Infinite Depth
Every organization, every team, every process can be decomposed into five roles. The pattern is fractal: each Executor becomes a Decision Maker at the next level down.
Decision Maker
Owns outcomes. Sets direction. Human at the top level.
Coordinator
Routes work. Maintains context. Where judgment concentrates.
Researcher
Gathers intelligence. Synthesizes options. Surfaces what matters.
Reviewer
Catches gaps. Validates quality. Embedded at every level.
The pattern repeats at every level. Click the Executor to see it again.
The Spectrum
Intelligence vs. Judgment
Not everything automates. The question is knowing which parts of your organization are intelligence (complex but rule-based) and which are judgment (experience, taste, strategic calls).
Intelligence
Complex but rule-based. Automates with AI.
Judgment
Experience, taste, strategic calls. Stays human.
Researcher
Mostly automates
Executor
Mostly automates
Reviewer
Interesting middle ground
Coordinator
Interesting middle ground
Decision Maker
Stays human
Researcher
Mostly automates
Executor
Mostly automates
Reviewer
Interesting middle ground
Coordinator
Interesting middle ground
Decision Maker
Stays human
The question isn't whether to adopt AI. It's knowing which parts of your organization are intelligence and which are judgment.
The Convergence
Why This Compounds
This isn't a one-time transformation. The organizations that build this muscle now will have a compounding advantage that's hard to replicate.
Today
Services
Human judgment combined with AI intelligence. We design and build the systems.
6 months
Compounding
More intelligence automates. Judgment focuses on highest-value decisions. Data from every engagement makes the next one faster.
18 months
Advantage
Institutional memory is deeper. Context architectures are more refined. Feedback loops have been running longer.
Their institutional memory will be deeper. Their context architectures will be more refined. Their feedback loops will have been running longer.