Five steps. In sequence. Non-negotiable order.
The Principle
Most AI implementations fail because the system was never mapped first. This methodology fixes that.
"The common mistake is reaching for a tool before the process is mapped. AI placed into an unmapped system doesn't simplify the system. It hides the chaos."
Map the System
Draw every step of the process as it actually exists, not how you think it works. Who does what, in what order, at what frequency. If you can't draw it, you can't improve it.
// most audits reveal 2–3 steps the founder didn't know existed
Find the Waste
Where is time lost? What tasks are repetitive, predictable, pattern-driven? Waste isn't just slowness — it's any step that a well-structured system shouldn't require a human to do.
// look for the steps that feel like they should be automatic but aren't
Apply the Filter
For each waste point: "Could I train someone to handle this and hand me back exactly what I need?" If yes, AI can probably do it. If no, the process needs more work first.
// the filter eliminates about 60% of apparent AI candidates
Place the AI
Only where it passed the filter. One process. One fix. Measure the result before adding anything else. This is the hardest step for founders who want to use everything at once.
// one well-placed AI tool outperforms four poorly-placed ones
Validate
Did the system get simpler or just different? Simpler means fewer steps, less human judgment required, less variance. Different without simpler means the AI is covering for a bad process — not fixing it. Remove it and go back to step one.
// if you can't measure it, you can't validate it
The AI Minimalist Diagnostic walks you through the five steps yourself. Free. Takes about 20 minutes.