Wondeur AI

Sophie Perceval and Olivier Berger built Wondeur AI on a contrarian reading of the art market: what looks irrational from the outside is a self-regulating system whose price movements can be described and predicted. On the TechTO stage in 2020 Berger put the stakes in balance-sheet terms — roughly US$1.8 trillion of art assets that mostly cannot transact, be lent against or be properly insured, because nobody can defend a value for them. The company's answer is to refuse auction data as the primary signal. Perceval and Berger argued from the stage that repeat auction transactions cover only about 3,000 of some 250,000 artists, and the company now says auction results are under 0.1% of the data needed to grasp an asset's value dimensions. Instead its models track endorsements, publications, movements and the influence networks of tens of thousands of market participants across more than a century. The current systems are branded WOND-A and sold to insurers and wealth managers rather than to collectors; the company says its client roster spans nine countries and two continents, that 85% of its spending goes to research, and that it has co-published research with UBS, Generali and Deloitte. It is headquartered at 215 Spadina Avenue in Toronto, with researchers in six cities across North America and Europe, and has run a co-creation programme it calls the Wondeur LAB since 2022.