Story · Hedge funds·
Contour Asset Management A/B tests AI agents across an earnings cycle
Hari Kumar of Contour Asset Management told AI Engineer how the fundamental long/short fund, with about 10 analysts covering 30 to 40 names each, documented an earnings cycle without AI and then re-ran it adding workflows one at a time, such as agent-run expert calls and agent-fielded consumer surveys. The firm tracks a metric it calls cost per useful insight. It pipes Yipit card data and AlphaSense transcripts into an internal platform over MCP and APIs instead of buying a bundled system like Rogo or Kensho, and Kumar said fixing connector breaks every few days still eats into the gains.
A rare first-person look at how a hedge fund measures whether research agents actually pay off, and at the build versus buy trade-off.
Source
StartupHub.ai · startuphub.ai · Found in the news · source checked Oct 25, 2026
Related tools
Spot an error? Submit a correction
More for Hedge funds
- SEC's FY2027 exam priorities list emerging technologies, including AI, alongside private funds and private credit
- LSEG agrees to acquire credit intelligence provider Octus for $3.535 billion
- Cognitive Credit raises $40M strategic growth investment led by Summit Partners
- Fiscal.ai puts estimates on nearly 10,000 stocks into its API and MCP server, including its own AI-built Agentic Consensus
One email a week. Free.