Odd Lots Goldman CIO Marco Argenti on the Warp-Speed Improvements in AI
274 snips
Mar 30, 2026 Marco Argenti, Goldman Sachs CIO and a key voice on AI in big finance, dives into how AI has gone from experiment to daily tool. He talks about coding agents speeding up developers, why firms are building more software in-house, how data access and info barriers shape adoption, and the balancing act between token costs, regulation, and moving fast without chaos.
AI Snips
Chapters
Transcript
Episode notes
Why Firms Want Engineers Directly From Model Makers
- Forward deployed engineers are product builders from model providers, not just support staff helping with integration.
- Marco Argenti says firms want direct access to the people building Claude or GPT because extra intermediaries slow learning during rapid change.
Employees Are Building Useful Tools In Hours
- Marco Argenti has seen employees use AI to migrate legacy on-prem applications to the cloud in hours.
- He also saw someone build a corporate travel assistant during a meeting by linking calendars, delays, and rebooking logic.
Centralize Model Routing To Control Token Costs
- Centralize model access through a gateway that routes each request to the best quality-cost tradeoff instead of letting teams call APIs freely.
- Marco Argenti says this reduces token anxiety, lets users stay creative, and keeps token spend below the value of the human hours saved.