An AI Lead Classification Pipeline for a B2B Jewelry Manufacturer

Nigaam Jewelers had built relationships with more than 250 jeweler accounts, and that was exactly the problem. Their sales team had no fast, consistent way to decide which accounts deserved attention first. Every account looked roughly the same on a spreadsheet, so reps fell back on gut feel and whoever happened to shout the loudest. High-value buyers sat untouched for weeks while time went into accounts that were never going to convert. The backlog kept growing, and the cost of guessing wrong grew right along with it.

We sat down with their sales and operations leads to learn how a strong account actually differs from a weak one, then turned that judgment into a structured, repeatable system. The result was a seven-phase pipeline built on Claude that reads each account, weighs it against the criteria that matter to their business, and assigns a clear classification in minutes rather than days. Each phase handles one job and hands clean data to the next, so the logic stays easy to follow and easy to adjust as priorities shift.

Classification is only useful if you can trust it, so we built inline validators into every phase. These checks compare each decision against the rules the team agreed on and flag anything that drifts, including the subtle misclassifications that their internal QA had been missing entirely. That safety net is what gave leadership the confidence to let the system run on its own instead of double-checking it by hand.

Today the pipeline classifies the full book of accounts automatically and ranks them so the sales team always knows who to call next. Work that used to stall in a backlog now happens quietly in the background, and the system has been running in production without intervention since launch. Reps spend their time selling to the right accounts instead of arguing about which ones those are.