“prediction markets failed at google because managers preferred control over accuracy”
Dan Schwarz, who built both of Google's internal prediction markets across two decades, tells the inside story of why they struggled and what it means for the future. Prophit (2005–2011) attracted 20% of Google employees but died when regulatory approval for an external launch stalled. Gleangen (2020–2022) reached over 10,000 users but faced a deeper problem: managers valued transparency and adjustability over raw accuracy, and the forecasts were asking internal questions when executives needed competitor intelligence on OpenAI and Microsoft. The piece closes with AI agents as the cost-reducing wedge that might finally make corporate prediction markets viable.
Some technical background helpful
Platforms mentioned: Polymarket, Kalshi, Manifold