Your AI Stack Isn't the Problem. Your Foundation Is.

Audit what you already have before adding anything new.

Issue one of The Friday Audit. The intro told you why this column exists. This is what it does.

I've watched the old model break in real time, and I've watched too many brands throw money at tools that assume the hard part is already done. I'm in the middle of the same shift, moving from wholesale into building AI systems for fashion brands, which means I get to say which parts work and which parts are bullshit, without softening it for the algorithm.

So let's get into what most brands between $1M and $10M are actually wasting money on with AI right now.

At $1M–$5M, the only job AI is genuinely useful for right now is auditing and pressure-testing what you already have. Most brands at this stage are adding new tools on top of foundations that are still weak.

The pattern I keep seeing at this level: brands buying new recommendation or personalization layers while 30% or more of their SKUs are still missing consistent fit notes, material specs, and occasion tagging. These tools will underperform and you will blame the software instead of the data quality.

Credit to the operators running real tests on their current recommendation logic with natural-language queries. If it fails on simple bundle or fit questions, you have found the actual constraint. Most brands skip this step and keep buying.

Run 10 of those queries this week. Real customer language, real fit questions, real bundle requests. The failures will tell you exactly where your catalog is lying to your tools. That is a free audit most brands never run.

At $5M–$10M the problem shifts again. You now have enough scale that weak data and inconsistent operations are actively limiting revenue, yet many brands are still treating AI as an add-on rather than infrastructure.

I'm watching this play out right now: conversational commerce and advanced styling features layered on product data that is still inconsistent. These features will underperform and you will blame the tool instead of the foundation it sits on.

Credit to the brands at $5M–$10M that are treating data quality like core infrastructure instead of an afterthought. They are the ones actually compounding gains from AI rather than just adding complexity and cost.

The gap between those two groups is not budget. It is sequencing. One group bought outcomes before inputs. The other fixed inputs and let the outcomes compound.

Stop adding layers. Start finishing the foundation. The brands winning with AI at this revenue range are the ones that made data and operational clarity the real work, not the tools that sit on top of it.