The AI bookkeeper files bank lines with built-in keyword rules before any learning happens. To find where they are thin, we write descriptors for a kind of business not yet covered and count what the rules do with them.
This round covered house cleaners, landscapers, yoga instructors, potters and dog trainers. Of 35 first-draft descriptors, the rules filed 18 right, 2 wrong and 15 not at all. We dropped five ambiguous lines (mulch delivery, a bare equipment repair, three deposits that could be sales or customer deposits). The real gaps were janitorial and clay supply houses, trade dues (ARS, NALP, CCPDT, Karen Pryor), pesticide-applicator and SoundExchange licences, and kiln repair. A janitorial supplier was wrongly filed as repairs and an arts-festival booth fee as platform fees.
After the rules, all 30 remaining descriptors file correctly. That is tuned on the same list, so it counts coverage, not accuracy. The descriptors are made up from public company names; no real bank export was used. Labelled real books are still what the project needs most.