We delivered an AI Provenance Engine for a Regional Fine Art Auction House

A fine art auction house handling six hundred lots per year was losing consignments to larger competitors because their provenance research took too long. Every piece that came through the door required a specialist to trace ownership history through gallery records, exhibition catalogs, estate inventories, prior auction results, and published raisonne volumes — work that took anywhere from eight hours for a well-documented piece to three weeks for something that changed hands privately in the mid-twentieth century. They had two researchers handling the load and a waiting list that pushed consignors to Christie's instead. Now an AI Provenance Engine processes incoming works against a cross-referenced corpus of auction records spanning forty years, digitized exhibition catalogs from over two hundred galleries and museums, published catalogue raisonne databases, estate sale records, gallery stock books that have been digitized from microfilm, import and export documentation archives, and insurance appraisal histories. When a consignor submits a piece it generates a preliminary provenance chain within four hours — identifying confirmed owners, likely owners with supporting evidence, and gaps that require human investigation. It flags authentication concerns automatically: stylistic inconsistencies cross-referenced against the artist's known periods, materials anachronisms based on technical analysis reports in the database, and provenance gaps that align with known forgery patterns or periods of conflict-era looting that require additional due diligence under current restitution frameworks. For the researchers this changed everything. Instead of starting from scratch with every lot they now begin with a structured chain that is seventy percent complete on average, focus their expertise on the gaps and judgment calls that actually require a trained eye, and clear pieces for catalog three times faster than before. The auction house reduced their average time-to-catalog from eleven days to three, eliminated their consignment waiting list entirely, caught two pieces with incomplete provenance that would have created legal liability, and increased their annual lot volume by forty percent without adding staff. Consignors who left came back because nobody else could turn around authenticated provenance research that fast.