Sotheby's uses machine learning tools to assist human appraisers in estimating prices for artworks, analyzing a database of over 50,000 repeat auction sales to generate market insights. The stated goal is to make appraisers more efficient, not to replace them.

Sotheby's internal machine learning team, led by VP Director of Machine Learning Ahmad Qamar and VP Director of Product Andrew Shum, has been developing systems to augment human appraisers by identifying comparable works through machine vision. The Sotheby's Mei Moses database — acquired in 2016 — underpins this work, providing repeat-sales data across eight collecting categories. For middle-market lots, the team sought to eventually automate price estimation and the end-to-end sales process, while for high-value lots the aim was to free specialists from repetitive tasks to focus on more complex estimates. The current status of full automation for any tier is unclear from available sources.