The framework

Build it right.
Then add AI.

AI isn't the problem. Stacking AI on top of broken, undefined processes and calling it a system is the problem. Every level in this framework builds on the one before it. Skipping a level doesn't save time. It creates expensive problems downstream that someone, usually you, has to fix manually.

The 5-Level Workflow Maturity Model

The goal isn't to stay at Level 1.
The goal is to earn each level before moving to the next.

Click any level, in the funnel or the list, to see how it fits.

01 · Manual02 · Automation03 · AI-Assisted04 · Hybrid05 · Agentic AI

Most businesses live at the top

  • 01

    Manual

    Most businesses belong here

    Do it manually first. Document every step. If you can't write it down, you're not ready to improve it.

  • 02

    Automation

    Most businesses belong here

    Once it's documented and repeatable, automate the rules. No judgment required. Most businesses should live here.

  • 03

    AI-Assisted

    Bring AI in only where human judgment is genuinely required. Nuanced decisions. Creative work. Novel situations.

  • 04

    Hybrid

    Solid automation at the base, targeted AI at the judgment points. Automation runs the rules. AI handles the calls that need reading between the lines.

  • 05

    Agentic AI

    Triggered, bounded, sitting on top of solid automation. Defined scope. Clear stopping conditions.

Higher levels use significantly more energy and compute. Solving something at a lower level is almost always the smarter and more responsible choice.

Download the Full Guide
The elephant in the room

AI uses energy. That's not nothing.

It uses real water too. The energy is real, the water is real, and the numbers matter. I'm not going to pretend otherwise to sell you something.

This is exactly why the framework exists. Responsible AI isn't a marketing label for me. It's why every level has to be earned before climbing to the next. If a documented SOP solves the problem, that's the answer. If automation solves it, that's the answer. AI gets used when AI is genuinely what the problem needs, not because it's the loudest tool in the room.

The answer isn't to never use AI. The answer is to be intentional about when you do.

It also helps to zoom out. Data centers aren't the only thing burning energy. Streaming an hour of video, leaving cloud storage running on files nobody opens, every push notification routed through somebody else's server. All of it adds up. We're not in a clean-energy economy. We're making trade-offs every time we touch a screen. AI is one of those trade-offs. A big one. But not the only one.

The most responsible thing you can build is a system that doesn't reach for the heaviest tool by default. Most days that means writing it down. Some days that means automating it. A few days that means AI. That order matters. That's the whole point.