98%+
Product-match accuracy, up from 80%
CASE STUDY · AI SAAS
An AI-native SaaS that worked, and read like a black box.
A confidential eight-person team built real AI for electrical distribution. The product functioned. But buyers could not see the value, and investors could not price the company. I rebuilt the commercial layer so the technology finally made sense to the people writing checks.
Electrical distribution looks like an AI problem until you get close. Then it is a messy data problem wearing an AI costume.
A distributor sells breakers, conduit, gear, and fittings across a catalog north of 300,000 items. The inputs are a mess. PDFs. Handwritten bills of material. Emails with half a part number. Vendor aliases for the same SKU. ERP records missing the attributes you actually need to quote.
The founders had trained a model that read all of it. Impressive. It also hid the real work. The hard part was never the language model. It was making 300,000 items legible enough that a machine could trust them.
This is the pipeline the whole business runs on. Raw distribution chaos on the left. A quote a rep can send on the right. Every stage exists to make the product data trustworthy before the model ever speaks.
The founders sold stage three. The value lives in stage one. My job was to make the whole line legible to a buyer and an investor, then price for it.
If the system cannot trust the product data underneath it, the LLM is just a confident intern with a calculator.
Every quote the system resolves teaches the catalog something. A new alias mapped. A missing attribute filled. A better alternative learned. The data layer gets sharper the more the product runs, and a competitor cannot copy that overnight because they have not done the resolving. The word here is moat, and it is earned, not claimed.
I paired that with pricing that finally matched the cost. This product burns real compute per customer. So I built usage-metered AI pricing that bills for the model cost it actually spends, per customer. Heavy users pay for heavy use, gross margin holds, and the unit economics stop leaking.
An AI product is not sold by its model. It is sold by the data underneath it and the story on top of it. This team had the hard technology. What they needed was a commercial layer that made the technology legible, defensible, and priced for the compute it burns. That is the Chief AI Officer job. Not the demo. The business around it.
The AI was never the hard part. The data was the whole job.