Everything Is a System

Complex Systems That Work Evolved From Simple Systems That Worked

That's John Gall's Law, and it's the most important principle I've learned about building anything that lasts.

You can't design a complex system from scratch and expect it to work. You have to start simple, make sure it works, then let it grow.

This principle matters more now because AI is fundamentally changing what kind of work humans do.

When AI handles 90% of electronic busywork, the remaining work shifts. You're no longer doing tasks. You're designing the systems that do the tasks.

If you can develop systems effectively, you build them once and they run forever. But most people never get there because they're stuck in the wrong type of work.

There's a difference between doing busywork and removing busywork.

If you're completing the actions required to finish busy work, you're doing busy work. If you're developing solutions to remove yourself and others from those processes, you're doing creative work.

You're building frameworks that require full understanding of the situation and how everything connects.

Start with the end in mind

I talk about inputs and outputs constantly. Knowing your desired inputs and outputs is essential. If you don't know what a good output looks like, designing your workflow will be nearly impossible.

Last week I wrote about the business process: see it, simplify, automate, scale it. That works for the whole structure. But there are fundamental components you need to understand:

• Error catching - Systems fail so build in ways to catch errors gracefully.

• Finished notifications - Know when something completes or breaks.

• Variable databases - Store the data that changes.

• Webhooks over database polling - Webhooks are quicker and more reliable than checking databases repeatedly.

• JSON - Computers read in JSON.

• Triggers, inputs, actions, outputs - The basic flow of any automation.

• Involved parties - Who needs to know what and when.

• Clean databases and pages as templates - Start with clarity, simplicity enables everything else.

Everything is a system. Every system is part of a larger system.

Why systems fail (and why that's normal)

I read Systemantics by John Gall a year or two ago. I just reviewed some notes from it yesterday. Definitely a timeless book about why systems work the way they do.

Some key insights:

System failure is intrinsic. Failure isn't a bug, it's a feature. Rather than blaming communication breakdowns, Gall argues that failure is inherent in systems themselves. It's behavioral inevitability and a good system handles failure gracefully.

New systems create new problems. When you establish a system to solve a problem, it generates additional problems unrelated to the original issue, purely by virtue of existing.

Example: You implement a CRM to solve "how do we track customer interactions?" But it introduces data hygiene requirements, integration maintenance, training overhead, and reporting disputes. New problems that didn't exist before.

Systems act up. They naturally misbehave, act unpredictably, and spend more time in failure mode than ideal operation. Murphy's Law applied to everything.

Gall said that most organizations blame communication when the real issues are behavioral properties of the frameworks themselves. Unintended side effects, inertia, and negative feedback.

Understanding Gall's principles helps you anticipate and solve these higher-order issues rather than being blindsided by emergent complexity.

What this means for building with AI

As AI automates busywork, humans shift from performing tasks to designing the systems that do them. This isn't just a job description change, it's a fundamental shift in required skills.

Gall's Law in practice means starting simple before building complex automated workflows. Having a structured process already, or first visualizing a new one, helps you begin simply.

Ensure your pages and databases are succinct and clear. That's the start of it all. Simplify the inputs to reliably get the desired outputs.

The original problems get solved: data entry, scheduling, basic decisions. But new ones emerge: managing continuous learning, adapting algorithms, monitoring cascading failures, reconciling different agent workflows.

These are systems problems, not task problems.

Leadership and strategy shift to designing resilient, flexible frameworks. Identifying weak links, slow adaptation, or failure modes only visible when agents interact. Evolving automation strategically rather than reactively.

The bottleneck is no longer code. It's systems thinking.

• Can you define workflows that evolve?

• Can you specify feedback and correction?

• Can you anticipate and design against systemantics failures?

As automation accelerates, mastering how systems work, not just how tasks are done, determines who thrives in a high-leverage business environment. Your understanding of frameworks is your lever for creating, managing, and evolving the new infrastructure.

Gall's warnings become practical survival skills: complexity breeds unpredictable failure, new systems spawn new problems, only evolutionary approaches endure.

To build and scale AI operations, you need a system for systems. Frameworks, protocols, and mindsets rooted in systems thinking, not just technical busywork.

Start simple. Make it work. Then grow it.

That's how you build anything that lasts.

Thanks for reading,

James